Full text of "Computer Language - Issue 06 (1985-02)(CL Publications)(US)"

See other formats

COMPUTER

$2.95

LANGUAGE

VOLUME 2, NUMBER 2

FEBRUARY 1985

C CHAIRMAN REPORTS
ON ANSI STANDARD

PJ. PLAUGER
LOOKS BACK

A UNIQUE LANGUAGE
CALLED MUMPS

TWENTY-ONE
COMPILERS COMPARED

SPECIAL C ISSUE

PSSST

JANUS/Ada

$99.95!!

PRESENTING THE NEW
JANUS/Ada C-PAKH

Janus/Ada Compiler

Janus/Ada Linker

Janus/Ada Libraries

Janus/ Ada Example/

Programs

Janus /Ada User Manual

AND THESE ADDED
FEATURES!!

1. Free User's Group

2. $99.95 Discount on
the Janus/Ada D-Pak

3. No License!!

4. No Copy Protection!!!

5. Customer tested for
over 3 years!!!

This is the introductory Ada™ package you've been waiting
for , . . over three years of actual field use, specifically on
microcomputers, by the government, Fortune 500 businesses
and major universities. Realistically priced, at $99.95, so you
can afford the most popular Ada implementation used on
microcomputers!

The new "C"-Pak is available for most microcomputers
running MS-DOS, including the IBM PC AT™, as are all the
other fine Janus/Ada programs. Call us or an authorized
distributor for your copy today!

Naiional Distributors

Internationa] Distributors

Westico, Inc.
25 Van Zant St.
NorualU. CI' 06855
(203) 853-6880

Trinity Solutions
5340 Thornwood Dr.
San Jose, CA 95123
(408) 226-0170

ASH II
7407 Marisol
Houston, TX 77083
(713)933-1828

A. O.K. Computers

816 Easley St., Suite 615

Silver Springs, MD 20910

(301)588-8446

Ada Australia
218 Lutwyche Rd.
Windsor 4030
QLD. Australia
(07) 57 9997

Microprogramming, Inc.
Suite 102 P.O. Box 3356

Chatsworlh, C A 91313
(818)993-6475

Progesco

155, rue du Faubourg

St. -Denis

75010 Paris

France

(I) 205.39. 47

Lifeboat. Inc. Japan
3-6, Kando-Nishikicho
Chivoda-Uu
Tokyo 101.. JAPAN
03-293-4711

ET M.C? M-». CCP K»»>rn»irni«iknilP**»*l K«""'
■ADA n ; iiarfrmaik of Th* V S Dirj«r.m« l i •* D*h**»
Mb DOS rl ■ indrmjik nf Mk)9H<i

OFTWARE, INC.

specialists in state of the art programming

P O Box 1512 Madison, Wisconsin 53701
(608) 244-6436 TELEX 4998168

CIRCLE 58 ON READER SERVICE CARD

PASCAIC COMPIf

ALA CART

Make Whitesmiths, Ltd. Part
of Your 1985 Software Strategy

For six years, software developers and systems integrators have looked to Whitesmiths Ltd
ror technologically superior compilers and multi-tasking operating systems.

Before you make your next move, contact Whitesmiths, Ltd.

Whitesmiths, Ltd.
97 Lowell Road
Concord, MA 01 742

TLX 95 1 708 SOFTWARE CNCM.

Whitesmiths, Ltd.

DISTRIBUTORS: Australia. Fawnray Pty Ltd.. Hurstvllle. (612) 570-6100; Japan Advanced
Data Con rols Corp ,. Chiyoda-ku. Tokyo (03) 263-0383; United Kingdom, Real T.me Systems
Douglas. Isle ol Man 0624-26021. Sweden, Un.sol, A B Gotebenj. 31-126810. * ys ' ems -

CIRCLE 85 ON READER SERVICE CARD

Squeeze More Out Of Every
On-Line Minute.

Presenting the software
package that makes your
computer more productive
and cost-efficient.

CompuServe's new Vidtex"" is compatible
with many personal computers sold today
(including Apple,® Commodore® and
Tandy/Radio Shack® brands). And it offers
the following features *-and more-to let
you communicate more economically
with most time-sharing services (including
CompuServe's Information Service).

Auto-Logon. Lets you log on to a host
simply and quickly by utilizing prompts
and responses defined by you. Also allows
qu ick transmission of p redefined responses
to host application programs after
logging on.

•Some versions of the Vidtex software do not implement all features listed

Vidtex is a trademark of CompuServe, Incorporated Apple is a trademark ol Apple Computet; Inc. Commodore is > trademark of

Commodore Business Machines. Radio Shack is a trademark uf Tandy Corp.

Function Keys. Let you consolidate
long commands into single keystrokes.
Definitions can be saved to and loaded
from disk file, allowing multiple definitions
for multiple applications.

Error- Free Uploading and Down-
loading. CompuServe 'B' Protocol con-
tained in Vidtex lets you transfer from
your computer to CompuServe and from
CompuServe to your computer anywhere
in the country. Also provides error-free
downloading from CompuServe's exten-
sive software libraries.

Full Printer Support. Printer buffer
automatically buffers characters until
printer can process; automatically stops
on-line transmission when full; and
automatically resumes transmission
when capacity is re-established. Also, lets
you print contents of textual video screen
or RAM buffer at any time.

Capture Buffer. Saves selected
parts of a session. Contents can be
written to a disk file; displayed both on and
off line; loaded from disk; and transmitted
to the host.

On-line Graphics. Integral graphics
protocol displays stock charts, weather
maps and more.

If you are already a CompuServe sub-
scriber, you can order Vidtex on line by
using the GO ORDER command. Other-
wise, check with your nearest computer
dealer; or to order direct, call or write:

CompuServe

P.O. Box 20212, 500O Arlington Centre Blvd.
Columbus, Ohio 43220

1-800-848-8199

In Ohio, call 614-457-0802
An H&R Block Company
CIRCLE 7 ON READER SERVICE CARD

COMPUTER

LANGUAGE

ARTICLES

The Standardization of C

by Jim Brodie

As a language well suited for both applications and systems program-
ming, C's popularity has prompted the development of a standard to
further enhance the language's portable syntax. Jim Brodie, chairman
of ANSI's C Programming Language Standards Committee, discusses
the history behind the standardization effort, the goals and working
principles of the committee, and how the language currently being
defined is different from the de facto, Kernighan and Ritchie standard.

C Instead of FORTRAN?

by Anthony Skjellum

Because of FORTRAN'S installed base of users and its limited ability to
be portable between hardware configurations, it has traditionally been
the default language for scientific and engineering programming tasks.
However, despite recent movements in the industry to give FORTRAN a
face-lift by adding certain structured language constructs, C may better
answer the need for an efficient, modular, and structured programming
environment for scientific and engineering projects.

Programming Macros in C, Part I

by Alexander 8. Abacus

In this three-part series, Alexander Abacus presents various methods
and techniques for using macros to improve C program development
efficiency. Part I focuses on how to use macros as building blocks for
parsing command line arguments. In the following two installments,
we'll see how macros can also be used in debugging large programs
and as a means for translating the common representation of data into
different but related tables required by different programs.

C to Assembly Interface

by Brian H. Burger

When programming large, complex projects, it often becomes neces-
sary to access architecture-dependent machine capabilities to increase
program speed and efficiency. Yet to some programmers, the C lan-
guage can open a Pandora's box since it requires more knowledge of
the machine than most other languages.

DEPARTMENTS

Publisher's Notes.
Feedback

CrossXthoughts

A new column on programming methods and techniques

Designers Debate.

UNIX on micros

ComputerVisions .

P.J. Plauger reflects on the history of C

Exotic Language of the Month Club

MUMPS: A multi-user, data management language

Product Bingo .

New product announcement column
Software Review

Special COMPUTER LANGUAGE C compiler analysis
Advertiser Index

104

Windows With A View
Toward The Future

The Window
Machine'" occupies
only 12K! Written in
tight, fast Assembler,
it performs like a
racing engine . . . with
more power than
you'll probably ever
need. Yet, it's an
engine designed to fit
in the vehicle of
your choice. ..from a
"stripped-down"
12BK IBM PC to a
fully loaded AT. The
programs you write
today will run on
the broadest range of
machines possible...
now. and in the
future.

Windows Bigger
Than Your Screen?

Here's where the
VSI part of our name
fits in, VSI means
Virtual Screen Inter-
face. Behind each
window, there's a
much bigger picture.
VSI defines virtual
screens rather than just windows. The
window itself shows whatever portion of
its virtual screen you wish to exhibit at
any given point in your program. Each
screen can be up to" 128 x 255 (columns x
rows, or rows x columns). And there are
more than 100 screen primitives at your
command.

Multilingual Windows

You can order The Window Machine
with the language interface of your choice:
C, Pascal, Compiled Basic, Fortran, Cobol,
or PLl. We've even recently completed

These are
coders'
windows...
designed to be
built into the
programs you
are writing.
They can
overlap, move
anywhere on
the screen,
grow, shrink,
vanish or blink.
They can be
bordered in
anything from
asimpieh'neto
flashing
asterisks. ..or
even no border
at all. And
you can have
up to 255 of
them at a time.'
CoJor or
monochrome
...of course!

Why did Simon &
Schuster, 3Com,
Tymshare, and
Revlon choose
VSI-The Window

figured if you wanted ribbons and bows
you could always add them yourself.)

And by offering you the product our-
selves, we were able to cut out all the
middlemen and save you a tremendous
amount of money.

The Window
Machine 1 "

/Kmlafjlt fur thi> IBM «:. XT, AT. IBM Compatibles,
nncl the Wans. T '- '"' l5ft

The Window Machine Includes:

and how come
you can buy it for
such a low price?

$59.95

• Zoom Windows
Multiple Virtual
Screens fup to 255J
i Choice of Borders
(including flashing borders]
■ Support for all Color and
Monochrome Video Attributes
(no graphics cord required)
Built-in Diagnostics
■ And much, much more

ORDER YOUR COPY OF

VSI— THE WINDOW MACHINE TODAY

For Visa and MasterCard orders call toll

free:

1-800 742-2500

an interface for Turbo Pascal*, so that
now true, full-featured windowing can be
utilized with this fine compiler. [Turbo's
own built-in "windowing" procedure is
extremely limited).

Windows That Won't Break You

We decided to save you a lot of money.
So. we left behind fancy binders, monn-
grammed slip cases and plastic pre-
sentation boxes. Instead, you'll find an
extremely powerful tool and a 200 page
manual written with an eye toward
simplicity, clarity and completeness. (We

'Tufoo Pascal is a Trademark of Borland International

The MiadoM Machine'' SS9 95 Shipping and handling included

LANGUAGE ICTEWACE:
□utllctC -Healia Cobol DMicrcsofl Basic Ccfnpikf U Micrasoll Fortran
T PLl CMttrosoll Pascal LTTurbo Pascal [fullteaturd true windotnriBl

COMPUTES

Citf

^IsU ?ip Hut.

□ ChetL CMoney Order

□VISA

MasterCard

Or! l

Pip Hjl*

'CaWormi* rKidenis lax included. Orders ojtsidt Ihe ISA: pi"** add Si
■■ i jhipp 1 r .'. ■■:■.: h .in-:' r: .

' AMBEB SYSTEMS
1171 S Saratoga-Sunnyvale Road
San lose, CASStiS

AMBER SYSTEMS, INC. 1171 S. Saratoga-Sunnyvale Road. San Jose CA 95129

TOR DEALER INQUIRIES: CALL OUR 80 D NUMBER

CIRCLE 2 ON READER SERVICE CARD

4 COMPUTER LANGUAGES FEBRUARY I9B5

Publisher's
Notes

' elcome to our
first theme
issue! This
month we spotlight the C programming
language.

The amount of reader interest and
increased popularity of C made the sub-
ject choice of this issue an easy one. The
C language has endured the evolutionary
process that any new language must
undergo when its followers attempt to
gather industry-wide acceptance for the
use of Ihcir language. After the past 15
years or so since its birth at Bell Laborato-
ries, C is regarded today as perhaps the
premier language for systems and applica-
tions programming because of certain
built-in features such as the portability of
source code, its simple but structured syn-
tax, and its machine-level interface
capabilities.

How much more popular will C be-
come? According to P.J. Plauger, involved
with the early development of C at Bell
Labs, C's popularity is just short of peak-
ing. (Sec editor Craig LaGrow's inter-
view with Plauger in ComputerVisions.)

A real highlight in this issue is our
comparative analysis of 21 C compilers—
nearly every C compiler now on the
market. This is the most complete and
objective (no re-run press releases here!)
C review any magazine has ever published.

We established a four-man team of C
experts who judged the compilers on
objective technical criteria, comparing the
compilers on the basis of four benchmark
timing programs and a close analysis of
their features. This consumer report kind
of approach seems to be the fairest
method of comparison possible, but I'm

sure we'll receive a healthy amount of dis-
agreement about the results. Let's just
keep it down to idle threats— no physical
harm, please!

Another important feature in this C
theme issue is a special report from the
chairman of the C Programming Lan-
guage Standards Committee, Jim Brodie,
on the history and current status of the C
language standardization effort.

This issue also introduces a new col-
umn by Namir Shammas entitled
CrossXthoughts. Look to this column for
lively discussions on topics like program-
ming techniques, algorithms, and prac-
tical solutions to complex programming
problems.

I also want to announce our plans to
sponsor a special seminar and workshop
on the C programming language to be
held in Cambridge, Mass., in September
1985. We're just setting up this event now
and would appreciate any suggestions you
might have. If you're interested in getting
more information about this special event,
please fill out the coupon on page 1 1 and
send it in today. Additional details about
the seminar will be released in upcoming
issues.

This issue is a mini celebration of
sons— it's COMPUTER LANGUAGE'S
half-year birthday. After only six issues,
we're selling over 30,000 copies every
month. Thanks for the great support.

Finally, remember COMPUTER
LANGUAGE is not a cheerleader for any
language. All Modula-2, Forth, COBOL,
BASIC, etc., fanatics, you'll have your
month too!

Cheers!

COMPUTER

Carl Landau

Publisher

Telecommunicate to COMPUTER LANGUAGE

COMPUTER LANGUAGE has established two bulletin board systems for you to
upload and download text and binary programs, as well as to leave your own elec-
tronic Letter to the Editor. All the program listings referred to in every issue of the
magazine will be available here.

In addition, COMPUTER LANGUAGE has its own Special Interest Group on Com-
puServe's national data base. After calling into your local CompuServe node, simply
type "GO CLM" al any prompt and you'll be in our SIG.

To access our bulletin board, set your computer or terminal to the following param-
eters: 8 data bits, no parity, I stop bit, full duplex, and either 300 or 1200 baud. The
telephone number is (415) 957-9370. After your modem makes the connection, type
RETURN several times, and everything else is easy.

Both systems are open 24 hours per day, 7 days per week. Due to the heavy number
of callers, please do not log into the system more than one time per day. Messages left
on either system will be combined the following day.

LANGUAGE

EDITOR

Craig LaGrow

MANAGING EDITOR

Regina Starr Ridley

TECHNICAL EDITOR

John Halamka

EDITORIAL ASSISTANT
Hugh Byrne, Lorilee Biernack!

CONTRIBUTING EDITORS

Doug Millison, Namir Shammaj, Ken Takara,
J.EdwardVolkstorfJr.

INDUSTRY NEWS CONSULTANT

Bruce Lynch

ADVERTISING SALES

Jan Denle

CIRCULATION COORDINATOR

Reriato Sunico

ART DIRECTOR

Jeonne Schacht

COVER PHOTO

Dow Clement Photography

PRODUCTION ARTIST

Anne Doering

PRODUCTION

Barbaro Luck, Steve Campbell, Kyle Houbolt

TECHNICAL CONSULTANT

Addison Sim$

MARKETING CONSULTANT

Steve Rank

ACCOUNTING MANAGER

Lauren Kalkstein

PUBLISHER

Carl Landau

COMPUTER LANGUAGE is published monthly by COM-
PUTER LANGUAGE Publishing Ltd., 131 Townsend St.,
San Froncisce, CA 94)07. (415) 957-9353,

Advertising: For information on ad rales, deadlines, and
piaceme/ir, contact Carl Landau or Jan Dente at (415}
957-9353, or write to: COMPUTER LANGUAGE, 13!
TownsendSt., Son Francisco, CA 94107.

Editorial: Please address all letters and inquiries to: Craig
LaGrow, Editor, COMPUTER LANGUAGE, 131 Townsend
St., Son Francisco, CA 94107,

Subscriptions: Contact COMPUTER LANGUAGE, Sub-
scriptions Dept., 2443 Fillmore St., Suite 346, Son Fran-
cisco, CA 94715. Single copy price: S2.95. Subscription
prices: S24.95 per year (U.S.J; 530. 95per year (Canada
and Mexico). Subscription prices far outside the U.S.,
Canada, and Mexico: S36.95 (surface mail,!, $54.95 (air
mail) — U.S. currency only. Please allow six weeks lor new
subscription service to begin.

Postal information: Second-closs postage rate is pending
at Son Francisco, CA and additional moiling offices.

Reprints: Copyright 1984 by COMPUTFP. LANGUAGE
Pub/ishingifd. All rights reserved. Reproduction of mate-
rial appearing in COMPUTER LANGUAGE is forbidden
without written permission.

Change of address: Please allow six weeks for change of
address to late effect. POSTMAST£R:Sendcbonge of ad-
dress (Form 3579) to COMPUTER LANGUAGE, 131
TownsendSt., San Francisco, CA 94107.

COMPUTER LANGUAGE is a registered trademark
owned by the magazine's parent company, CL Pubiica-
tions. ANmoteriQl published inCOMPUTER LANGUAGE
is copyrighted C 1984 by Ct Pub/icafions, Inc. AH rights
reserved.

NEW from BORLAND!

TURBO TOOLBOX & TURBO TUTOR

4KMtwtE

\ \~^

"TURBO is rauc_
Pascal IBM sells."

Jerry Pournelle,
Byte, July 1984

"TURBO PASCAL appears to violate
the laws of thermodynamics.

You won't find a comparable price/
performance package anywhere. It
is simply put, the best software deal
to come along in a long time. If you .
have the slightest interest in
Pascal. . .buy it."

Bruce Webster,
Softalk IBM: March 1984

y x

3?..

BORLAND

INTERNATIONAL

GIFT PACK

$9995

A SAVINGS OF $30!

What a gift for you and your friends! The extraordinary TURBO PASCAL
compiler, together with the exciting new TURBO TOOLBOX and new TURBO
TUTOR. All 3 manuals with disks for $99.95.

PASCAL Version 2.0 (reg. $49.95). The now classic program
development environment still includes the FREE MICROCALC SPREAD SHEET.
Commented source code on disk

• Optional 8087 support available for a small additional charge

NEW! 1 OLBOX (reg. $49.95). A set of three fundamental

utilities that work in conjunction with TURBO PASCAL. Includes:

• TURBO-ISAM FILES USING B+ TREES. Commented source code on disk

• QUIKSORT ON DISK. Commented source code on disk

• GINST (General Installation Program)

.Provides those programs written in TURBO PASCAL with a terminal installation module
just like TURBO'S!

• NOW INCLUDES FREE SAMPLE DATABASE . . . right on the disk! Just compile
it, and it's ready to go to work for you. It's a great example of how to use TURBO
TOOLBOX and. at the same time, it's a working piece of software you can use
right away!

NEW! 1 TUTOR (reg. $29.95). Teaches step by step how to use the TURBO

PASCAL development environment-an ideal introduction for basic programmers.
Commented source code for all program examples on disk.

30 DAY MONEY BACK GUARANTEE Available at your nearest software dealer.

For VISA and MASTERCARD order call toll free: l-(SOO)-255-800S l-(800)-742-1133

(Lines open 24 his.. 7 days a week) Dealer and Distributor inquiries welcome (408) 438-8400

CHOOSE ONE (please add $5.00 for handling and shipping U.S. orders)

All Three-Gift Pack $ 99.95 + 5.00 SPECIAL! Turbo Toolbox $49.95 + 5.00

All Three & 8087 139.95 + 5.00 SPECIAL) Turbo Tutor 29.95 + 5.00

Turbo Pascal 2.0 49.95 + 5.00 Turbo 8087 89.95 + 5.00

Check Money Order . VISA . MasterCards

Card #:

My system is: 8 bit 16 bit

Operating System: CP/M 80 CP/M 86 MS DOS PC DOS

Computer: Disk Format: ^___

Please be sure model number & format are correct.

NAME:

ADDRESS: . ^__

CITY/STATE/ZIP:

TELEPHONE:

Exp. date:

Shipped UPS

California residents add 6% sales tax. Outside U.S.A. add $15.00 (if outside of U.S.A. payment must be by bank draft payable in
the U.S. and in U.S. dollars! Sorry, no CO.D. or Purchase Orders. „„

t BORlflflD

9) INTERNATIONAL

4113 Scotts Valley Drive
Scotts Valley, California 95066
TELEX: 172373

CIRCLE 6 ON READER SERVICE CARD

>«tec The Most Powerful C

for the IBM AT • MACINTOSH • MS DOS • CP/M-80 • ROM APPLICATIONS
IBM PC/XT • APPLE // • CP/M-86 • TRSDOS • CROSS DEVELOPMENT

Why Professionals Choose Aztec C

AZTEC C compilers generate fast, compact
code. AZTEC C is a sophisticated development
system with assemblers, debuggers, linkers,
editors, utilities and extensive run time libraries.
AZTEC C is documented in detail. AZTEC C is
the most accurate and portable implementation
of C for microcomputers. AZTEC C supports
specialized professional needs such as cross
development and ROM code development.
MANX provides qualified technical support.

AZTEC C86/PRO
— lor the IBM AT and PC/XT

AZTEC C86/PRO provides the power, portabili-
ty, and professional features you need to
develop sophisticated software for PC DOS, MS
DOS AND CP/M-86 based microsystems. The
system also supports the generation of ROM
based software for 8088/8086, B0186, and 80286
processors. Options exist to cross develop ROM
code for 65xx, 8080, 8085, and Z80 processors.
Cross development systems are also available
that target most micro computers. Call for infor-
mation on AZTEC C86/PRO support for XENIX
and TOPVIEW.

POWERFUL — AZTEC C86/PRO 3.2 outper-
forms Lattice 2.1 on the DHRYSTONE
benchmark 2 to 1 for speed (17.8 sees vs 37.1)
while using 65% less memory (5.8k vs 14k). The
AZTEC C86/PRO system also compiles in 10%
to 60% less time and supports fast, high volume
I/O.

PORTABLE — MANX Software Systems pro-
vides real portability with a family of compatible
AZTEC C software development systems for PC
DOS, MS DOS, CP/M-86, Macintosh, CP/M-80,
APPLE // + ,//e, and He (NIBBLE -4 apple rating),
TRSDOS (80-MICRO - 5 star rating), and Commo-
dore C64 (the C64 system is only available as a
cross compiler - call for details). AZTEC
C86/PRO is compatible with UNIX and XENIX.

PROFESSIONAL — For professional features
AZTEC C86/PRO is unparalleled.

• Full C Compiler (803618086 ■ 80186 ■ B0286I

• Macro Assembler lor 808818086(80186/80206

• Linkage Editor with ROM support and overlays

• Run Time Libraries - object libraries + source
DOS 1.x; DOS 2.x; DOS 3.x; screen I/O; Graphics;
UNIX I/O; STRING; simulated float; 8087 support;
MATH; ROM; CP/M-86

. Selection of 8088(8086. 80186. or 80286 code genera-
tion lo guarantee best choice for performance and
compatibility

• utility to convert AZTEC object code or libraries to
Microsoft format. {Assembly + conversion takes
less than half the time as Microsoft's MASM to pro-
duce MS object)

• Large memory models and sophisticated memory
management

• Support products lor graphics, DB, Screen, & ...

• ROMablecode + ROM support + separate code and
data + INTEL Hex Converter

• Symbolic Debugger & Other Utilities

• Full Screen Editor (like Vi)

• CROSS Compilers are available to APPLE //, Macin-
tosh, CP/M-80, TRSDOS, COMMODORE C64. and
ROM based 65xx, and 808O/BOB5/Z8O

• Detailed Documentation

AZTEC C86/PRO-AT $500

(configured for IBM AT - options lor 8088/8086)

AZTEC C86/PRO-PC/XT $500

(configured tor IBM PC/XT . options for 80136/80286)

AZTEC C86/BAS includes C compiler (small model only),
8086 MACRO assembler, overlay linker, UNIX, MATH,
SCREEN, and GRAPHICS libraries, debugger, and
editor.

AZTEC C86/BAS $199

AZTEC C86/BAS(CP/M-86| $199

AZTEC C86/BAS (DOS + CP/M-86) S299

UPGRADE to AZTEC C86/PRO S310

C-TREE Database with source $399

C-TREE Database (object) .$149

CROSS COMPILERS
Cross Compilers for ROM, MS DOS, PC DOS, or CPIM-86
applications.

VAX - > 8086/BOxxx cross S5000

PDP-11 ->8086/80xxx cross $2000

Cross Compilers with PC DOS or CP/M-86 hosts are $750
lor the first target and $500 tor each additional target,
Targets: 65xx; CP/M-80; C64; 8080/8085/ZBO; Macintosh;
TRSDOS; 8086/80B8/801 86/80286; APPLE //.

AZTEC C68K

— for the Macintosh

For power, portability, and professional features
AZTEC C68K-C Is the finest C software development
system available tor the Macintosh.

The AZTEC C6BK-C system includes a 68000 macro
assembler, a linkage editor, a source editor, a mouse
based editor, a SHELL development environment, a
library of UNIX I/O and utility routines, lull access and
support of the Macintosh TOOLBOX routines, debug-
ging aides, utilities, make, diff, grep, TTY simulator with
upload & download (source supplied), a RAM disk (tor
512K Mac), a resource maker, and a no royalty license
agreement. Programming examples ar included. (Over
600 pages of documentation).

AZTEC C68K-C requires a 128K Macintosh,
and two disk drives (frugal developers can make
do with one drive). AZTEC C68K supports the
512K Macintosh and hard disks.

AZTEC C6BK-C (commercial system) $500

AZTEC C68K-p (personal system) $199

AZTEC C68K-P to AZTEC C68K-C upgrade $310

Mac C-lree database $149

Mac C-tree database with source $399

Lisa Kll (Pascal to AZTEC C68k object converter) . .$ 99

AZTEC C65

— lor the APPLE //

"...The AZTEC C-system Is one ot the finest software
packages I have seen..." NIBBLE review. July 1964,

The only commercial C development system available
that runs native on the APPLE II + , He, and lie, the
AZTEC C65 development system includes a lull floating
point C compiler compatible with UNIX C and other
MANX AZTEC C compilers, a 6502 relocating assem-
bler, a linkage editor, a library utility, a SHELL develop-
ment environment, a full screen editor, UNIX I/O and
utility subroutines, simple graphics, and screen func-
tions.

AZTEC C65 (Apple DOS 3.3) $199

AZTEC C65/PRO (Apple DOS + ProDos) $350

(call for availability)

AZTEC C ll/PRO

— lor CP/M-80

The first member of the AZTEC C lamily was the
CPiM-BO AZTEC C compiler. It is "the standard" com-
piler for development on CP/M-80. The system includes
the AZTEC CMC compiler, an 8080 assembler, a linkage
editor, an obiect librarian, a full library of UNIX I/O and
utility routines. CP/M-80 run time routines, the SMALL
library (creates modules less than 3K In size), the fast
linker for reduced development times, the ROM library,
RMAC and M60 support, library source, support for
DRI's SID/ZSID symbolic debugger, and more,

AZTECC ll/PRO $349

AZTEC CII/BAS $199

C-TREE Database with source $399

C-TREE Database in AZTEC object form $149

AZTEC C80

— lor TRSDOS (Radio Shack Model HI * 4(
•7 ve had a lot ot experience with ditterent C compilers,
but the Azlec C80 Compiler and Professional Develop-
ment System is the best I've seen." 80-Micro. Decem-
ber, 1984, John B. Harrell lit

This sytem has most of the features ot AZTEC C II lor
CP/M. II is perhaps the best software development
system lor the Radio Shack Model III and IV.

AZTEC CB0 model 3 (no floating point) $149

AZTEC CB0 model 4 (lull) $199

AZTECC80/PRO(lulltormodel3and4) $299

To order or for information call:

MANX ®

.ill

1-221-0440

(20 1) 530-7997 (NJ and outside U.S.A.). Or wrlle: MANX
SOFTWARE SYSTEMS, P.O. Box 55, Shrewsbury. N.J
07701.

SHIPPING INFORMATION ■ Standard U.S.
shipment Is UPS ground (no fee), In the U.S.
one day shipment is $20. two days is $10.
Canadian shipment is $10. Two days ship,
ment outside she U.S. is by courier and is
freight collect.

TRS B0 RADIO SHACK TRS DOS is a trademark Of TANDY.
APPLE DOS MACINTOSH is a trademark ot APPLE.

For Technical Support
(Bug Busters) call: 201-530-6557

CIRCLE 69 ON READER SERVICE CARD

FEEDBACK

AC limerick

Dear Editor:

C is one of the most versatile languages
available today. Its ability to add by func-
tion allows you to create large applica-
tions. Yet for operating systems or pro-
cess control, C can "get small" while still
maintaining good structure.

As a mildly typed language, C
impresses some restrictions on the pro-
grammer but still allows freedom for spe-
cial situations. I thought your readers
might enjoy a little program I wrote for
nonprogramming managers who insist on
easily readable code (Listing 1).

Mike Rejsa
Minneapolis, Minn.

Correct Ada figure

Dear Editor:

For those people who want a correct
copy of Figure 2 of my article "Learn to
Think in Ada" (Nov. 1984, pp. 47-49),
here is a copy (Listing 2). Please apolo-
gize to our readers for me. I can imagine
how frustrating it must be to try to figure
out how a program works when it isn't all
there.

Do-Whi!e Jones
ftidgecresl, Calif.

*/
*/

/* A limerick for C programmers

/* by Mike Rejsa

#define a

Idefine and(c) bdos(6, c)

#define created

fdefine Dee =

fdefine had =

#define it —

#define thought puts

#define with -

int *fame =128, *give =128, *name =128;

int *Bea = 130, *looking = 130, *to = 130;

int language = 120;

main()

{

/* A lady programmer named */

*Bea Dee

created a language

with 'C;

while ( — *looking)

{

for (*fame had *to; *give; it a *name)

and( ' » );
thought ( "Is it Bea, C, or Dee?" );

Listing 1.

What was it, exactly?

As many readers have pointed out, the
listings in Joe Celko 's "Wiiat Day is ft,
Exactly" (Dec. 1984, pp. 47-50) contained
some errors. Here are corrections, by list-
ing and line number:

■ Listing 1/line 4: change MOD 400 to
MOD 100

■ Listing 2 /line 1: change Validate to
ValidDale

line 4: change MinSystemYear to

MaxSystemYear

line 11: change MonthSuze to MonthSize

line 12: change second MinSystemYear to

MaxSystemYear

line 13: change Validdate to ValidDate

■ Listing 3 /line 2: add */ to end of line
line 8: change IN1T90 to IN1T (0

line 9: change month > 1 to month > 2

with C0NS0LE_I0; use CONSOLEJO ;
procedure Solve_Exercise_3__4__B is

I, J : integer;

TEST_RESULT, I_IS_0DD, J_IS_EVEN : boolean;
begin

put( "Please enter an integer ""J"" "

put("Now enter another integer ""J""

if I mod 2 = 1

) ; get(I); new_line;
"); get(J); new_line;

then I_IS_0DD ;

else I_IS_0DD ;
end if;
if J mod 2 =

then J_IS_EVEN

else J_IS_EVEN
end if;

TEST RESULT := I_IS_0DD and J_IS_EVEN;
put ("The test turned out ") ; put(TEST_RESULT) ; new line;
end Solve_Exercise_3__4_B;

Listing 2.

TRUE;
FALSE;

:= TRUE;
:= FALSE;

IF (raonth=2) THEN Jul2:= Jul2 + 1 ELSE NULL;

IF ((month>2) AND (month<9)) THEN Jul2:= Jul2 - 1 ELSE NULL;

Listing 3.

IF LeapYear(year) THEN X:=l ELSE X:=2

Julian:= Truncate(275 * month / 9) - X * Truncate((month+9) / 12)

Julian := Julian + day - 30;

Listing 4.

For vour IBM/PC

mbp COBOL:

4 times faster;
and now witn

SOKT& CHAIN

I75Q

mbp COBOL can be
summed up in one
word: fast.

Because it generates
native machine language object code, the
mbp COBOL Compiler executes IBM/PC*
programs at least 4 times faster (see chart)

GIBSON MIX Benchmark Results

( lit ulated S-Proflk'
tKepresentaltiv COBOL statement mix)

Execution lime ratio

mbp Leve! II"

COBOL COBOL

LOO

1.08

R-M"*
COBOL

5.98

Microsoft*
COBOL

6.18

allow source & object
code, map & cross-
reference checking; GSA
Certification to ANSI '74
Level II; mbp has it all.

It's no surprise companies like Bechtel,
Chase, Citicorp, Connecticut Mutual, and
Sikorsky choose mbp COBOL; make it
your choice, too, mbp is available at
Vanpak Software Centers, or direct.
For complete information, write mbp
Software & Systems Technology,
Inc., 7700 Edgewater Drive, Suite
360, Oakland, CA 94621, or phone
415/632-1555
-today

12SK syMem wltfi hard disk required "[BM'PC is in IBM TM: "Level [I
is a Micro Focus TM. ""A Ry^n-McrarUrulTM. *■■■* Microsoft TM.

Fast also describes our new SORT; which
can sort four-thousand 128-byte records in
less than 30 seconds. A callable subroutine
or stand-alone, 9 SORT control fields can
be specified. And our new CHAIN is both
fast and secure, conveniently transferring
control from one program to another, pass-
ing 255 parameters. Plus, new extensions to
ACCEPT & DISPLAY verbs give better, faster
interactive programming.

The complete COBOL. An Interactive
Symbolic Debug Package included standard;
Multi-Keyed ISAM Structure; listing options

CIRCLE 39 ON READER SERVICE CARD
10 COMPUTER LANGUAGE! FEBRUARY 1985

line 12: change day: to day:

■ Listing 4/line 2: change * to */
line 7: change 30. 475 to 30. 43 75

line 12: change month > 1 to month > 2

■ Listing 5 /line 8: change
CenturyTahk'l 1:121 to CenturyTahlel 1 :5J
Line 26: change : to : at the end of line

■ Usting6Vline 12: change EFto IF.

Calendar modifications

Dear Editor:

Lines 8- 1 1 of Listing 4 in Joe Celko's
"What Day is It, Exactly?" are wrong.
They might be replaced with the two lines
in Listing 3.

An easier approach to the Commercial
Dale is shown in Listing 4.

F. Barry Mulligan
Atlanta, Ga.

Leap year rule

Dear Editor:

I feel I should correct an inaccuracy in
Joe Celko's "What Day is It, Exactly?"
The inaccuracy has to do with leap year
calculations.

First, in the text of Celko's article he
states, "... every year divisible by four
is a leap year . . . there is also a 400-year
cycle . . . The four-year cycle exists
because the number of days in a year is
365 and almost one-quarter days.
Leap year takes care of the one-quarter
day, and the 400-year rule takes care of
the 'almost.' "

The actual rule is different. The year is
a little more than 11 min short of 365 days
and one-quarter. This amounts to being a
day ahead after about 128 years or about 3
days ahead in 400 years. The rule is: if the
year is divisible by 4, it IS a leap year,
UNLESS, if the year is divisible by 400, it
IS a leap year. Hence, the year 2000 is a
leap year.

Daniel Efron
St. Louis Park, Minn.

Author Joe Celko comments: Where
365. 25 days per year appears , you can
substitute the more accurate 365.2422
days per year if the software will not have
rounding problems in the calculations.

PROGRAM Math

PARAMETER (a=l. 43576, b=5. 23417)

total = 0.0

DO 10 i = 1, 2500

c = EXP( L0G( (SQRT(b))**2 ) )
d = ATAN( SIN(c)/GOS(c) ) /a
total = total + d

10 CONTINUE

WRITE(*,20) total

20 F0RMAT( IX, 'Total is: \F11.5)

STOP

END

a/b

Listing 5.

PROGRAM Overhead

PRINT *, 'Finished'

STOP

END

FORTRAN math program

Dear Editor:

Your software review of the Digital
Research FORTRAN-77 compiler (Nov.
1984, pp. 66-71) was worthwhile but it
lacked any critical comments. Being
interested in such a compiler, I visited a
friend who recently purchased the DRI
product and is running it on an NEC/APC
operating under CP/M 86. We entered the
two programs in Listing 5.

The resulting statistics for program
"overhead" are provided: (most of the
run-time is consumed in reading the
executable image into RAM)

Compilation time: 41 sec
Link time: 5 min 24 sec
Run time: 16 sec
Executable file size; 84K

The math program was compiled
without options and then linked with the
HARD8087 option (requires 8087
coprocessor) and also with the SIM8087
option (software mathematics), and these
two executable images were run from a
RAMdisk. The following statistics are
provided:

■ On the 8086 math program, an
:xecution time of 15 min 15 sec, and a
result of 2499. 99976

■ On the 8087 chip, an execution time of
15.7 sec, and a result of 2499.99976.

Note that the correct result is 2500.

Kenneth C. Beaudrie
Denver, Colo.

ANNOUNCING . . . COMPUTER LANGUAGE'S

C Seminar/Workshop

Cambridge, MASS ■

September 1985

Plans are being set now for COMPUTER LANGUAGE'S C
Seminar. The 2/2 day event will be held in the fall of 1985
in beautiful Cambridge.

Details concerning the topics, speakers and dates wiil be
announced soon. This seminar will be the most
comprehensive and practical session ever held about the C
programming language.

Become involved in this event from
the start by filling out this — ■.*"■■* " \

c ° upon ,oday: :;"»" *■ -*" wd '

bout'

Send to: COMPUTER LANGUAGE Seminar
l 3 1 Townsend Street • San Francisco, CA 94 1 07

WARNING:
Do NOT read this flowchart:

it UP '

■.cltir' i -*

_«»

—

■ unotlirr 1 ■ -i. *

■,

:..fesr..:

i Y"»

" CtrtAtn MMAI *

„•* pid -.^

iifillT

Att^* to * T«

riTw

Unless you have time to spare ... in extreme oases a few people have found
they had thirty years to spare.

You read the flowchart anyway. Why? Because flowcharts are a powerful

graphic way of communicating ideas. The big problem is producing them:

they take a lot of time to do well and are difficult to revise or correct without

starling over from scratch.

The flowchart above was produced using EasyFlow, a computer aided

flowchart generation tool. You decide how the flowchart is to be laid out

and describe the flowchart to EasyFlow using a simple command

language. EasyFlow then does the hard part of actually producing and

printing the flowchart.

EasyFlow is a well designed, thoroughly tested and comprehensively

documented package,

• Fast. EasyFlow is written in assembly language for speed. It produces
a typical flowchart in 12 seconds.

• Easy.The EasyFlow flowchart description language is straightforward
and easy to learn.

• Printers. EasyFlow works with all printers. Extensive printer control
facilities allow you to fine-tune EasyFlow to produce the best possible
flowcharts on your particular printer.

• Shapes. EasyFlow comes with eighteen standard flowcharting shapes.
User defined shapes can be easily added.

• Lines. User definable line-drawing characters allow EasyFlow to adapt
to personal preference and printers with graphic line characters.

• Manual. Complete, comprehensible and over 100 pages long. Also in-
cluded is a quick reference card, a planning grid and ten demo
flowcharts (demo flowcharts supplied on disk as well).

(U.S. funds) Check, money order, Visa or company RO.

HavenTree Software Limited

$49.95

EasyFlow is available for MS-DOS/PC-
OOS machines (96K or greater) on IBM
PC format 5" diskettes. Available for
CPfM-80 machines (48K or greater) on 8"
SSSDand most soft sectored 5" formats.

R.R. #1, Suite 100
Seeley's Bay, Ontario
Canada, KOH 2N0
(613) 542-7270

CIRCLE 41 ON READER SERVICE CARD
12 COMPUTER LANGUAGES FEBRUARY 1985

MULTI-BASIC

The Compatible BASIC Compiler from Alcor
(Supported Features Chart)

JFELSE/FQfiNEXT
WH.LEWEN0
OPEN /CLOSE

•

•-
ft

'* ' *

* ft

• ft

CREATE/DELETE
PRINT/ PRINT USING
LPRINT/LPAINT USING

•
•

• *

ft •

PRINT @
INPUT/LINE INPUT

input (t/une input K

• .

• ft

BEAD #/READ# LINE
LOC/IOF/EOf/ERROR
FIELD/GET/PUT

•

• ft

RSET/LSE17ERASE
MKD$/MKfS/MK5$
CVi/CVS/CVD/SPC

READ/DATA/ RESTORE

TROM/TROFF

OEFFN/RAN00M/RND

:•■■ ■ *

DEF USft/SWAP/WAIT
CALL {ASSEMBLY LANG)
DEFSTR/DBL/3NG/INT

DOUBLE/REAL

INTEGER/STRING

G0TO/GOSUB

*
•
•

ON ERROR SOTO
RESUME/RESUME NEXT
EJtL/ERR

ON number GOTO/GOSUB

NAME/ftENAME

PEErOPOKE/mP/OUT

*
*
•

•
*

SYSTEM/SOUND

SADD/MATCH/UCASES

VAL/TAB/STRS/VARPTR

*
•

STN/C0S/TAN/ATN
10G/EXP/ABS/SQR
COMMANDS/IF END

•
•

ft
•
•

INKEY$/INPUT$

TIMES/OATE5/HEXV0CTS

STRWGS/SPACES

*
*

LEFTS/RSHTS/MIDS
CHRS/ASCfLEN/SGN
OPTION BASE

*
•

* ■
*

* *

R0W/POS/LPOS

DIM/MEM/FRE

M0D/MFRE

•

• •

GET/PUT (CHARACTER)

KILL/CLS/ClEAR/fNSTR

mT/RX/CtNT

•
*

. * *

CSNG/CD8L

CHAIN

C0N5TABWCONCHAR%

■

*
ft

• ft

* •

C0NS0LE/LPRINTER
BINARY RANDOM FILES
ASCII RANDOM FILES

•

*
•

* «

• •

255 CHARACTER NAMES
REDIWENSfONED ARRAYS
UNLIMITED STRING SIZE

* - *

MULTI LINE FUNCTIONS
PROCEDURES/RECURSION
FUNCTION TRACING

•

ft ft

LOCAL VARIABLES
NESTED FUNCTIONS
OPTiONAL LINE NUMBERS

*
*

* •

• ft

DESCRIPTIVE LABELS
SINGLE PRECISION
DOUBLE PRECISION
LINK TO PASCAL &C

•
•

■

*
•

• ft

For TRSBO models I, II, III, 4, 12, or 2000 using TRSDOS,
CP/M, or MSDOS and lor IBM PC using PCDOS
Multi-Basic, Pascal, or C $139
Add shipping $6.00 USA, $28 Overseas

ALCORB

13534 Preston Road. Suite 365
Dallas. Texas 75240
(214)238-8554

'.'i.i-.-Ei.ii.L n, .1 trademark a! Alcor Systems

TflSDOS TRSBO a a registered liademarli ol Tandy Corporation

CP/M. CBAS'C are i'aaemart.5 g! Digilal Research

MSDOS MBASIC are Haaemartj q! Mhcreiofi

CIRCLE 1 ON READER SERVICE CARD

crossXthoughts

Pseudocode syntax

By Namir Shammas

leonte to
CrossXthoughts!
It is a

new column for all programmers and a
platform for sharing algorithms and pro-
gramming development techniques and
solutions.

We will cover a wide variety of sub-
jects, such as data management tech-
niques, interpreters, graphics, and arti-
ficial intelligence. 1 will be using the
COMPUTER LANGUAGE Bulletin Board
Service (415 957-9370) and CompuServe
to interact with readers (but ordinary mail
is not completely ruled out!). These media
will be used for topic selections, follow-
up discussion on published columns,
questions asked by readers, and sharing
expert opinions and experiences.

In addition, the column will assist read-
ers with similar interests in contacting
each other. This will allow further dis-
cussion and exchange and published
follow-up. I will maintain SIG lists con-
taining names, addresses, phone num-
bers, etc. They will be posted on the bul-
letin boards and also made available
through COMPUTER LANGUAGE by
written requests. Putting your name on a
particular SIG list is done at your request,
by writing or through the BBS and Compu-
Serve. The column will also carry out
mini-contests awarding books, software
packages, and subscriptions in the
magazine.

o make this col-
umn a crossroad
for programmers
who use different languages, I will be
using pseudocode listings frequently.
Pseudocode is also this issue's topic.

To insure good readability and a degree
of consistency I am suggesting a syntax
for the Programmers' Pseudo Listing
(PPL). I have loaded it on a BBS file
called NILE, which contains an applica-
tion program written in Turbo Pascal and
the documentation for using it. It checks
the PPL syntax and helps in top-down
software design process allowing devel-
opment from general to specific.

Additional advantages gained from
jsing the PPL are:

■ During software development it is eas-
cr and faster to alter and update text than
it is to change flowcharts.

■ The PPL offers programmers flex-
ibility in sharing information. One can
generalize certain code segments to either
spare the other party irrelevant code or
deliberately hide detailed algorithms to
protect programming tips and tricks.

■ To assist top-down software design.

■ To assist applications development in
more than one language or from one to
another.

The implementation of a standard PPL,
just like any language, is subject to certain
syntax and rules. They are of course much
more flexible than those for formal lan-
guages. Deviating from them is by no
means a sacrilege. The general rules are:

■ Data types and identifiers. This is the
area where languages differ greatly. The
PPL will leave the choice of how

to present data structures up to the
programmer.

■ Labels. The aim of structured lan-
guages is to move away from using labels
and GOTOs. It is worth pointing out that
Modula-2 has not implemented the GOTO
statement at all! By contrast, Ada and C
allow : using the GOTO especially for
exiting deeply nested loops. Therefore,
labeled GOTOs are allowed. This should
benefit those who program in BASIC and
FORTRAN. However, I must say-
following the example of nonsmoker
signs— "Thank you for not GOTOmg]"

■ Loops. There are two basic types of
loops. The first has a loop counter with

General loop construct

starting and ending values— the code in

the loop is repeated a fixed number of
times. This is the famous FOR-DO loop in
Pascal and Modula-2 or the FOR-NEXTin
BASIC.

The second loop type would execute the
enclosed code depending on the success of
a conditional test. When the test is located
at the beginning of the loop, we have the
WHILE-DO, as in Pascal, Modula-2, C,
and FORTRAN 77. When the test is car-
ried out at the end of the loop, forcing
execution of the code at least once, we
have the REPEAT-UNTIL effect. The test
can also be carried out anywhere inside
the loop in an attempt to exit. This is
equivalent to the LOOP-EXIT'm Ada and
Modula-2.

From the preceding we can see that the
different loop types can be presented
basically as shown in Listing 1. The /A7-
TIALIZE will remind the programmer of
any loop initializations; I suggest its pres-
ence be mandatory. The word "None"
should be used when no action is needed.
The LOOP keyword is followed by an
optional loop name. The latter is useful to
keep track of exiting nested loops (actu-
ally, it is an Ada feature).

The loop BEGIN may contain informa-
tion to indicate the use of a counter or a
conditional test to simulate the WHILE-
DO construct. The conditional test can be
inserted inside the loop code or at the end.
The latter will simulate the REPEAT-

INITIALIZE <optional set-up>

LOOP <optional name>

BEGIN <no text to simulate REPEAT-UNTIL> or

<declare loop counter> or

<0ptional test here to EXIT <optional name> >

loop code with optional test to EXIT <optional name>

Optional test here to EXIT <optional name>

END LOOP

TERMINATE resolve any pending operations

Listing 1.

UNTIL construct. The END LOOP token
need not be followed by a loop name since
we expect loops to be logically nested.
The TERMINATE key word is used to indi-
cate the action taken to resolve pending
data treatment after the loop ends.
■ Conditional branching. Inspired by
languages like C, Pascal, Modula-2 and
Ada, I use two types of conditional bran-
ching constructs:

1 . The well-known IF-THEN construct
with ELSE and ELSEIF clauses for more
elaborate decision-making. Listing 2
shows the two alternatives when using the
IF-THEN clauses.

2. The CASE test is applied when an
identifier can be used as a switch to select
the appropriate course of action. Listing 3
shows the construct for CASE . The
WHEN and OTHERWISE keywords will
indicate the beginning of a new CASE
option and the end of a previous one,
except for the first WHEN.

IF-THEN constructs

IF <Test

is true>

THEN

Outcome

# 1

ELSE

Outcome

# 2

END IF

and

IF <Test

1 is true>

THEN

Outt

;ome

# 1

ELSEIF <Test

2 is

true>

THEN

Outcome

# 2

ELSEIF <Test

3 is

true>

THEN

Qut<

;ome

# 3

ELSE

Outi

sortie

# 4

END IF

Listing 2.

■ Procedures, and functions. Due to the
use of plain English words to describe the
action taken, we must be able to point out
specific calls to procedures and functions.
I suggest that the procedure names be
enclosed in brackets and function names
in braces. The following example shows
how a call to a procedure SOLVE is
documented:

Set up simulataneous equation and
[SOLVE]

Another example is a call to a logical
function DiskFull :

IF (DiskFullJTHEN warn the user that
no space is left

Since the PPL may be involved in early
software development stages, the proce-
dure and function declarations follow the
main code portion. This allows for a top-
down software design. Thus, procedures
and functions that are called more often
would be located further down in the list-
ing. To declare the code for a procedure
or function, it is sufficient to use the key-
words PROCEDURE or FUNCTION fol-
lowed by a name.

■ Input/Output. This is another aspect of
programming where different languages
as well as dialects implement differently.
We will adopt the following commands:

DISPLAY < list > for screen output
INPUT < list > for keyboard input
PRINT < list > to send output to a

printer.

File I/O involves reading and writing of

data records. The first command needed

by all file I/O is to open it:

OPEN < filename > , < buffer name or
number >,< mode >

where the I/O mode is one of the
following:

INPUT for data input

OUTPUT for data output

CASE construct
CASE <identifier>

APPEND to append output data to the
existing file

RANDOM for random data read and
write.

Its counterpart is CLOSE < buffer name
or number > to close the file.

For sequential I/O we have READ
< buffer > , <data > to input data,
WRITE < buffer > , < data > to output.
For random I/O we have READ
< buffer > , < record # > , < data > and
WRITE < buffer > , < record M > ,
<data > to read and write, respectively.
The SEEK < buffer > , < record > is used
to position record pointers to a specific
location.

Port communications is carried out in a
similar way to file I/O except the modes
allowed are, obviously, INPUT and
OUTPUT.

■ Incrementing, decrementing and
scaling variables. These operations are
used extensively in all languages. Some,
like C, offer a shorthand way of stating
them. I am adopting the C notation. This
is useful in shortening code lines since
long variable names would most likely be
used for clarity. To increment a variable:

Variable + = Added number

which is equivalent to

Variable = Variable + Added number

and to decrement a variable:

Variable - = Subtracted number

Similarly, scaling a variable by multi-
plication or division, as in

Variable = Variable * Factor

and

Variable = Variable / Factor

become (Variable *= Factor) and (Vari-

WHEN <value 1 is ture> => Outcome # 1
WHEN Oalue 2 is ture> => Outcome # 2

OTHERWISE => Outcome # X

SND CASE

Listing 3.

14 COMPUTER LANGUAGE ■ FEBRUARY 19B5

able/ = Factor), respectively.
■ Remarks. For those who want to for-
malize remarks and comments, borrowing
from Ada's syntax is useful. The remark
is expressed by at least two consecutive
dashes, as in;

— This is a remark

This one too!

and so is this —

Listing 4 shows pseudocode for solving
a single nonlinear equation using the
Newton-Raphson method. The PPL
shown reflects a rather detailed code
revealing very specific algorithms.

Listing 5 shows the code for a
regression program that will read data
from a file and perform a best-fit
regression. This is done by subjecting the
data read to a combination of mathe-
matical transformations. Each of the latter
represents a different regression model.

The code shown will process the data
through four models. The slope, intercept
and correlation coefficient is calculated
for each model. An insertion sort is per-
formed as results are obtained for each
model. This allows for the printing of the
curve fitting information sorted from best
to worst model. The PPL listing reflects
the case where a programmer is avoiding
detailed coding in certain areas.

l o you have any
I comments or
suggestions
concerning PPL? Does the idea of a con-
sistent type of pseudocode appeal to you?
Is it too restrictive? Would you like to par-
ticipate in a SIG-PPL group for further
refinement of PPL? Let me hear from
you. Please write to me in care of
COMPUTER LANGUAGE, 131 Townsend
St., San Francisco, Calif. 94107 or con-
tact me through the BBS or CompuServe.

In the next issue I will discuss inter-
active parsers and interpreters for mathe-
matical equations (i.e., the heart of elec-
tronic spreadsheets). If you have any code
or algorithms to share, questions to ask,
or would like to enrole in the SIG list,
drop me a line. Remember that the SIGs
are there to help exchange information
and, who knows, perhaps find your next
software coauthor! H

PPL listing for solving a single nonlinear equation
using the Newton-Raphson method

PROGRAM Root

Version 1.0 June 10, 1984

LAST UPDATE : 06-15-84 12:22:23

INPUT Guess, Accuracy, Maximum

INITIALIZE Set Counter =
Set Goflag True
LOOP

— The token below is not followed by anything as part of

— simulating the REPEAT-UNTIL loop
BEGIN

IF abs(Guess) > 1 THEN Increment = 0.01 * Guess

ELSE Increment = 0.01

END IF

Delta_X = 2 * Increment * {F(Guess)} /

({F(Guess + Increment)} - {F(Guess + Increment)]

Guess -= Delta_X

Counter += 1

IF Counter > Maximum THEN Goflag is False END IF

When (abs(Delta_X) <= Accuracy) OR (Goflag is False) 'EXIT

END LOOP
TERMINATE None
IF Goflag is True THEN

DISPLAY Guess and Counter
ELSE

DISPLAY Message for divergence
END IF

END PROGRAM — This is optional

FUNCTION (F(X)}
— User function

F = EXP(X) - 3 * X * X

END

END PPL

Listing 4,

X X K H

PRENTICE-HALL:

THE C/URIIX*

C TITLES

1. Programming in C with a Bit
of UNIX*, F. Richard Moore. A first
glimpse of C for the computer novice.
1985. 388 pp., paper [73009-3} S 16.95;
cloth |730IO-1) S22.95.

2. Learning to Program In C, Thomas
Plum. Plum walks programmers through
the world of C. 1984, paper (52784-6)

S 2500; cloth (52785-3) S32.95.

3. C Programming Guidelines,
Thomas Plum, Put yourself on solid
ground with this clear explanation of
the standards. 1984. 145 pp.. paper
(10999-1) $26.00.

4. The C Programmer's Handbook.
AT&T Bell Laboratories. M. I. Bolsky.
Destined to be the C programmer's
constant companion. 1985. 96 pp..
spiral (11007-2| S 14.95.

5. The C Programming Language
Brian W. Kernighan and Dennis M.
Ritchie. The "Bible'— a must for C pro-
grammers. 1978, 228 pp., cloth
(IIOI6-3| S22.95.

6. C: A Reference Manual, Samuel
P. Harbison and Guy L. Steele. The
latest details of C— right at your fmger-
tipsl 1984, paper (11000-7) $19.95; cloth
(11001-5) $24.95.

7. The C Puzzle Book. Alan R.
Feuer. Challenge your mind with these
C brain teasers. 1982. paper (10992-6)
$14.95; cloth (10993-4| S 18.95.

8. Comparing and Assessing
Programming Languages— Ada,
C. and Pascal. Alan R. Feuer and
Narain Gehani. Which language suits
you best7 Feuer and Gehani spell it outl

1984, 256 pp.. paper (15484-9) $16.95;
Cloth (15485-6) S24.95.

UNIX TITLES

1. The UNIX* Programming
Environment. Puts two UNIX experts
right at your side. 1984. paper (93768-0)
SI9.95: cloth (93769-6) $26,95.

Brian Kernighan and Rob Pike

2. A UNIX* Primer, Ann Nicols Lo-
muto and Nico Lomuto. Lays out UNIX
by cutting through common -problems.

1983. 240 pp., paper (93773-0) $20.00;
cloth (93888-6) 5 21.95.

3. UNIX* for People. Peter Birns.
Patrick B. Brown, John C. Muster. Learn
the UNIX language in your language!

1984. 544 pp.. paper (93744-1) S 19.95:
cloth |93745-8| 29.95.

ORDER YOUR FREE 15-DAY

TRIAL COPIES TODAY!

By phone: dial (201) 767-9520 or

write to Steven T Landis, Prentice-Hall

Book Distribution Center. Route 59 at

Brook Hill Drive. West IMyack. New York

10995 OR . . .VISIT YOUR LOCAL

BOOKSTORE

"UNIX is a trademark of AT&T Bell Uborarones

PPL listing for best-curve fit

PROGRAM BEST_FIT

Version 1.0 June 10, 1984

LAST UPDATE : 06-15-84 12:22:23

INPUT Filename — Get name of data source

OPEN Filename, #i, INPUT

READ #l,Number_of_data — Read the amount of data

INITIALIZE None — Loop to read the (X,Y) data pairs

LOOP

BEGIN For counter = 1 to Num_of_data
READ #l,X(counter),Y(counter)

END LOOP

TERMINATE CLOSE #1 — close buffer

Set Numjfodel = 4 — This sets number of models
[DO REGRESSION]
Print .sorted results
END PROGRAM

PROCEDURE [DO REGRESSION]

INITIALIZE None

LOOP Outer

BEGIN For model = 1 to Num_Model

INITIALIZE Set regression summation to zero
LOOP Summmations

BEGIN For count = 1 to Num_of_data
[Transform]

Update summations using X and Y
END LOOP Summations
TERMINATE None
Calculate Slope, Intercept and correlation coefficient

— Perform insertion sort using the values of the

— correlation coefficient
[SORT]

END LOOP Outer
TERMINATE None

END

PROCEDURE [Transform]

CASE model
WHEN 1

WHEN
WHEN
WHEN

END CASE

END

END PPL --

=> X = X(count); Y = Y(count)

=> X = Log(X(count)); Y = Y(count)

=> X = X(count); Y = Log(Y(count))

=> X = Log(X(count)); Y - Log(Y(count))

This must be included in every PPL listing using
program NILE. PAS

Listing 5.

CIRCLE 83 ON READER SERVICE CARD

16 COMPUTER tANGUAGEB FEBRUARY 1985

DESIGNERS DEBATE

UNIX on micros

NIX has been

around on mini-
computers for a
long time. This month, however, let's
focus on the future of UNIX as a micro-
computer operating system alternative.

Many factions in the microcomputer
industry have stated that UNIX may never
receive the market acceptance UNIX sup-
porters claim it will primarily because of
certain inherent problems with the UNIX
operating system's overall organization
and interface to the programmer and
end user.

Let's turn now to a discussion of UNIX
on micros and focus on certain arguments
that deal with UNIX as a development
environment.

The following people debate the UNIX
issue in this month's column: John
McNamee, a free-lance consultant: Jack
Crenshaw, software engineer; Bob Peter-
son of the MUSUS SIG on the Compu-
Serve data base; Mike Cohen: Morton
Goldberg; Tim Smith; Bob Upshaw;
Earle Robinson; Eli Willner; Joel Swank;
and Dennis Hamilton.

s it reasonable to
run UNIX on a
micro? It has a
reputation as a rather large operating sys-
tem, requiring vast amounts of memory
and disk space.

McNamee: Define micro and you
will get a better answer. I doubt a
Commodore 64 will run it. and the IBM
PC has a hard time. My view is that UNIX
requires a CPU made by somebody other
than Intel and a disk drive that is faster
than the ST506 interface. 100-millisec
seek drives you find in PCs.

Peterson: I have had occasion to
look into how much machine UNIX
needs. I came to the conclusion that 256K
RAM. memory management/protection
and 5MB of hard disk space are the abso-
lute minimums for UNIX.

Certain clone systems— for example.

By Ken Takara

IDRIS— require less disk space and no
memory management. This places UNIX
in the medium to upper ranges of micros
at a usual cost of S5 ,000 th rough $7 ,000
for the basic machine. This would be for a
single-user machine.

I tend to believe that UNIX is a better
fit with the architecture of the MC68000
and National 32000 processors. In the
Intel world, an 80286 is probably the real-
istic minimum for decent performance.

Cohen: 256K RAM? Try doing any-
thing useful on a 256K TRS- 16! 5 1 2K is
more like it. That and a 5MB hard disk
would still be a very tight fit. A nice thing
would be a streamlined, single-user
UNIX subset.

McNamee: That's not really true.
The PDP-1 1 is a mini, and it can only
address 64K of code plus 64K of data, and
you can't even play tricks to get more.
There are no user-accessible segment reg-
isters. The definitions of super-micro,
mini, super-mini, and mainframe seem to
constantly change. When the PDP-1 1/70
came out, everybody said it was almost a
mainframe. Now you can get an
LSI-1 1/73 with about the same speed right
on your desk.

I think Fortune tried to sell floppy-
based UNIX. You will notice that their
profits (or should I say, lack of them) may
have had something to do with trying to
sell systems too small to run UNIX.
Tandy threatened to release floppy-based
XENIX, but so far it has not shown up. I
don't know anybody else doing it,

Goldberg: I'll stick my two cents in.
I had VENIX/86 running on my PC-XT I
killed it because it wanted too much disk
space. It would only allow PC-DOS 3MB
of the 10MB on the hard disk. My
PC-DOS applications outgrew 3MB and
VENIX had logo.

VENIX/86 is a real UNIX, an AT&T
licensed version 7 with some Berkeley
enhancements such as the vi screen editor.
Code developed under VENIX should run
on most UNIX systems if it is recompiled
on the system it is moved to. I got VENIX
to develop code for what I thought was
going to be the booming UNIX-on-PCs
market. But the market didn't boom.

I earn more money from stuff I do on
PC-DOS than I could with the VENIX
stuff. So it was an economic decision that
forced me to abandon VENIX. If I could

have afforded a second hard disk, I would
have kept them both.

Crenshaw: It seems to me every-
one's missing a key point: UNIX is a time-
sharing system. That means that one CPU
is used to poll or otherwise respond to a
usually large number of user terminals. It
also means a very considerable overhead
is required to allow that CPU to do all the
bookkeeping necessary to support that.

Thai means context switching, access
control, virtual memory overhead, prior-
ity control, accounting, mail and message
handling— not to mention the ubiquitous
news! — etc., etc. That doesn't leave much
CPU left over for number crunching, as
anyone who's ever used a 40-user VAX
can testify.

UNIX is without a doubt the best OS
ever written for multi-user, time-sharing
systems, bar none. Now the question is.
do such systems have any place on a
micro? As a strong supporter of Purnelle's
law— at least one CPU per user— 1 claim
they do not. For some reason, whenever a
new computer is announced (for example,
the IBM PC AT) the first question asked
is, will it run UNIX? The next one is. how
many users will it support?

My question is. why do we want lo
crowd people onto a single CPU? Has sil-
icon suddenly become a precious metal?
To get back to sinale-user UNIX, isn't
that" OS -9?

McNamee: There is no reason why
UNIX can't be run single user. The only
difference between multi-user and single-
user UNIX is an entry in /etc/iTys. A pro-
cess is a process is a process. It doesn't
matter if they are all attached to the same
terminal or to multiple terminals.

1 think the reason everybody asks if it
runs UNIX and how many users it sup-
ports is that networks aren't doing the job
they should. Right now it costs less to
have one machine serving multiple users
than it docs lo network a bunch of PC's.

Peterson: UNIX Is one of the few
operating systems that will live com-
fortably on a local area network. A single-
user UNIX machine in a LAN satisfies the
morc-than-onc processor/user criterion
while providing multitasking. OS-9 is not
UNIX, nor is it a single-user UNIX.

There is a need in [he real world to
share data. At the current time, a multi-
user machine is the best understood envi-
ronment for data sharing. The LAN envi-
ronment is finally getting some software
packages that share across the network,
but they are few.

At the present time, a 3 to 10-user
machine is cheaper than 3 to 10 micros.
This is because of the mechanical devices
and packaging, not the silicon . When the
economics of the world change so that a
LAN environment with data sharing tools
is less expensive than a multi-user
machine, and such a system is available
from a trusted vender, then LAN systems
will become more prevalent than multi-
user systems.

Hamilton: I believe that UNIX hav-
ing been implemented in a time-sharing
setting is irrelevant to a discussion of what
UNIX is. There isn't much about the
architecture that users see that commits
UNIX one way or the other.

You build it with a time-sharing kernel
when you've got a large facility that you
want to share. When you have small facil-
ities that you want to interconnect, you do
it differently. But the user, apart from

observed performance differences, should
be able to get work done on either with the
same knowledge and skills.

ow, who would
want to use the
world's most

impossible system . . . ?

McNamee: A programmer with an
IQ greater than his or her shoe size should
have no problem. Users are best given
menu shells or another such interface.
Turning a dumb user over to one of the
standard shells is cruel and unusual
punishment.

UNIX was a system designed by pro-
grammers for other programmers. It does
make it very easy to set a user up with a
menu shell, but you need a programmer to
do the initial setup. I think the same could
be said about CP/M or MS-DOS. The
CP/M prompt A> doesn't give you much
more to go on than + or % .

Peterson; UNIX is an old system.
Many times this means that there is a lot
of software available, and for UNIX this
is true. But in the UNIX world, this soft-
ware isn't oriented to the business or per-
sonal user but to the academic or software
developer.

The fact that UNIX was designed in the
early 1970s also means that it was based
on contemporary design principles and
hardware. This implies that UNIX itself
doesn't know about bit-map displays, for
example. Some applications do. and some

venders do ship bil-map-oriented user
interfaces, but these are add-ons. not a
part of the basic system.

The average user will be better served
with TopView on an IBM PC AT or with
some other system for which appropriate
applications software is available.

U pshaw: The only way to address
the severe shortage of good software is to
make it possible to write a piece of soft-
ware once and use it on a large number of
different machines.

It's clear to even the most casual
observer that UNIX is running on more
different machines than any other oper-
ating system. It's only natural that UNIX
migrate to micros.

Crenshaw: I think when many peo-
ple talk about UNIX, they are really talk-
ing about the user interface. And that can
begotten in a lot of OS's. As a matter of
fact, I think wc should be sure we are all
talking about the same aspects of the OS.

The part of UNIX that 1 think is least
suited to micros is the file system and the
terminal interface software. I love the
directory structure and the shell.

McNamee: 1 am a programmer. I

Hard problems?

Csharp can help!

Cut your development time with C source code for realtime data acquisition and
control. The Csharp Realtime Toolkit includes: graphics, event handling,
scheduling, and state systems. Processor, device, and operating system
independent.

SYSTEMS
GUILD

Systems Guild, Inc., P.O. Box 1085, Cambridge, MA 02142
(617) 451-8479

CIRCLE 89 ON READER SERVICE CARD

18 COMPUTER LANGUAGE! FEBRUARY 1985

know what I'm doing. I usually get things
right. UNIX is for people just like me. It
provides me with the tools to get my job
done in the least amount of lime with the
least effort. It also allows me to easily cre-
ate any user interface I desire.

I can set up a machine that secretaries
can use without ever knowing about any
UNIX commands. AM they need to do is
log in. The only people who lose with
UNIX are the ones who need help (i.e.,
dumb end users) but don't buy a system
with proper support fin the form of a pro-
grammer to set them up).

It is very important to remember that
UNIX is not MS-DOS or CP/M. It is a
multi-user operating system. I don't know
of any good multi-user system that is so
simple a total novice could make it go
without help. Sure, it's easier to use
MS-DOS, but MS-DOS does not solve the
same problems as UNIX.

Smith: I like UNIX too, as a pro-
grammer's system. But I curse it many
times, for example, when I forget some
bizarre aspect of. say, shell syntax. Also,
as a one-time UNIX system administrator
on a big VAX, I know how bad it can be
when you have a bunch of dumb and/or
unfriendly users aboard.

We need something better. There's no
reason why it can't be done. Maybe it's
mice and icons for the end users and a
good shell for the experts, both on the
same system. But it's going to have to be a
lot more efficient than any current UNIX.

Robinson: I can certainly agree that
UNIX is better than CP/M, which is only
a very primitive file loader. Bui to say that
it is better than MS-DOS for a single user
is something else. UNIX, due to its con-
stant file accessing, would be unaccept-
ably slow on most micros unless it were a
CPU of the MC68000 family.

Otherwise, with the features now avail-
able under MS-DOS and the excellent
public domain utilities— like directory
programs, extended type-ahead buffer,
NDOSEDIT, editors, squeeze and
unsqueeze programs, shells,
etc. — MS-DOS is the system of prefer-
ence for most people other than those
actually doing program development .

McNamee: Does dir really mean
more that As? I doubt it. Most of the places
where I've seen a large number of dumb
users doing computer work, they all had
their instructions written down, and they
just followed the steps. They really didn't
know what they were doing, and in this
kind of environment, it doesn't matter
whether you type delete , era , or rm to
erase a file.

Crenshaw: Count me in as one who

has been less than overwhelmed by
UNIX. Well, to be more accurate, the
thing that turns mc off most is not so much
the system as the aura of mystery that
seems to surround it. It's what I call the
Fraternity Syndrome.

In brief, a young computer science stu-
dent gets his first exposure to UNIX in
school. He's totally blown away by it and
totally in awe of the UNIX wizards. But
after stumbling around for a few years, he
learns some things and wakes up one
morning to find that he's one of the wiz-
ards. He's held in awe by others and that
feels pretty good.

Now do you think he's going to tell you
all you need to know to use UNIX effec-
tively? No way. You're going to have to
pay your dues, same as he did.

I find that this attitude pervades every-
thing about a typical UNIX installation,
including the command names and the
documentation. Tome, it takes a lot of the
fun out of things. It's like some kind of
gigantic inside joke.

Willner: One more vote against
UNIX. I don't like to fight my environ-
ment. I don't like to memorize archaic
commands, no matter how much it makes
me look like a wiz. I don't like to destroy
all my files because of a misplaced
keystroke.

Swonk: UNIX is hard to learn, but
once learned, it is very powerful. I can do
in one line in UNIX what would take me
200 cards in IBM-360JCL.

Pet erson: Willner pointed out a cou-
ple of the reasons I don't like UNIX. The
command names don't relate to the func-
tion of the command and I. too, don't like
an environment where dangerous oper-
ations don't prompt for verification.

However, my dislike goes deeper.
UNIX doesn't have a consistent method of
record locking, and the file system is
excessively fragile. The OS doesn't take
advantage of the fact that almost all termi-
nals are now CRTs, not TTYs. Not to
mention that UNIX doesn't understand
bit-mapped displays.

One of the great myths about UNIX is
that there is a single thing named UNIX.
This just isn't so. User groups are just
now defining what they consider the core
of UNIX, including OS calls and library
utilities. In addition to the various ver-
sions delivered by AT&T and Berkeley,
each processor's and/or vender's imple-
mentation differs in subtle ways.

I like the tools philosophy embodied in
UNIX and the concept of pipes. However,
the idea of pipes deserves to be extended.
It hasn't really changed since UNIX was
first written up in SIGPLAN notices.

Upshaw: The UNIX user interface
allows one to easily go from one machine
to another. Sure VisiCalc will port fairly
easily from one CP/M machine to another
(say, an Apple to a Xerox 820), but it's a
rewrite to port to an IBM PC, not to men-

tion a VAX or a Cray (yes. there is a
spreadsheet on the Cray). And. of course,
the UN1X/C environment is conducive to
portability.

McNamee: I like UNIX as a pro-
grammer's environment and would hate to
see it turned into a pure end-user system. I
think it is quite possible to build a reason-
able interface on top of existing UNIX. I
don't think you can build a programmer
friendly interface on top of a Mice/Icon
system.

In its intended environment, the Mice/
Icon interface works well. It just doesn't
do anything good for me as a program-
mer, and in fact, it holds me back. I want
to bring up again my point that it isn't fair
to compare the user friendly aspects of
UNIX to existing single-user OS's.

Compare it to other time-sharing sys-
tems. Multi-user systems such as UNIX
were created with goals very different
from those of the Mac. UNIX is good for
what it was designed, and the Mac is good
for what it was designed. UNIX is not the
OS for everybody or even for the majority
of the people.

Upshaw: When I am interviewing
candidates for a job, more often than not
they have used UNIX. UNIX is very pop-
ular in the universities, and with all the
major venders jumping on the UNIX
bandwagon, its popularity is likely to
increase.

When I hire someone, I don't want to
train my employee to use five different
systems— I've got more important things
to do. If he or she already knows the sys-
tem that runs all five of my machines. I'll
save a considerable amount of time and
money. ■

Smith: Try to get hold of a copy of
the November issue of UNIX Review .
There's an interview with Don Norman, a
professor of psychology at Univ. of Cali-
fornia at San Diego. In it he explains why
he does and doesn't like UNIX. He's a
sharp guy and has become something of
an expert in the design of systems for
good human interaction.

eriodically, I will
be holding argu-
ment clinics over
some subject on the CLM SIG on Compu-
Serve. If you don't happen to have a
CompuServe account, you can si ill join
in. I will try to announce the future sub-
jects in advance on the COMPUTER
LANGUAGE Bulletin Board Service, so
send your thoughts through the mail or via
the BBS. II

THE MAGAZINE
FOR THE FULL-OUT

BRAINO.

Computer Language is written for
people who can program in Two
or more computer languages.

Let's face it, that leaves out most
people. Programming is a rigor-
ous, intellectual discipline and
Computer Language magazine
is the first and only publication
dedicated exclusively to this field.
Your source for the latest
technical skills and methods
used by software specialists.

We cover the major develop-
ments in the software design field,
from theory to implementation.
Computer Language focuses
on the most important and
useful language design /\
information available in
the fast-moving micro- i
computer industry.
Written for the
person who takes
computing
seriously.

We're talking
about you — the
experienced
software author,
programmer, or
engineer who
routinely programs
in two or more
high-level
languages.

A person who understands the
creative nature of programming
and appreciates the beauty of
efficient code in action.
COMPUTER LANGUAGE will
constantly challenge your
abilities.

The foremost industry experts will
discuss: • Algorithmic Approaches
to Problem Solving ■ Language
Portability Features • Compiler
Designs ■ Utilities ■ Artificial

20 COMPUTER LANGUAGE ■ FEBRUARY1985

COMPUTER »

LANGUAGE-

V. ,

Intelligence ■ Editors ■ New
Language Syntax ■ Telecommuni-
cations • Language Selection
Criteria ■ Marketing Your Own
Software • Critical Software &
Hardware Reviews

Send lo;
COMPUTER

LANGUAGE

2443 Fillmore Street Suite '346 San Francisco, CA 941 1 5 1

YES! Start my charter subscription to

Computer Language. My I year charter
subscription is just $24.95, a 33%
savings under the single copy price,

a $24.95 D Bill me.

Paymenl Enclosed

Name

Address

City

State

Zip

COMPUTERVISIONS

PJ. Plauger reflects on the history of C

c calls himself a
programmer al
heart. But as an
accomplished science fiction writer, pres-
ident of a successful software company,
technical book author, musician, and run-
ner, P.J, Plauger is certainly a man of
many talents.

His friends and work associates know
him as Bill — a name he was affectionately
given by his older sister three days after
his birth. But in any published, profes-
sional work, he prefers to use the initials
of his real name, Phillip James.

Plauger "s first contact with Bell Labo-
ratories was as a undergraduate at Prince-
ton Univ. in the early 1960s. He and
several of his fellow students were fortu-
nate enough to shuttle back and forth
between Bell Labs and school and get
involved in temporary summer projects.

"In those days it was heaven," says
Plauger, "It was the place to go because
just about anything you could do that
would advance the state of communica-
tions in the next five years was considered
fair game. And it was the place to be if
you were interested in being at the cutting
edge of programming."

"In the 1960s, there was still this tre-
mendous intellectual freedom in the very
undefined world of computer program-
ming," he remembers. "Maybe a place
like IBM Yorktown Heights. RCA, GTE,
or XEROX would come close, but Bell
Labs was really just a wonderful place to
be for freedom and access to resources."

Plauger went on after Princeton to earn
a Ph.D. in nuclear physics from Michigan
State Univ. There he spent half his time,
with a fellow graduate student, designing
and writing a virtual memory operating
system. He confesses that he "invented"
demand paging, not knowing that the rest
of the world had already done so.

"I've worked and earned my living as a
programmer since 1963. and I say that
very proudly." says Plauger. "I consider
myself a programmer rather than a man-
ager or a computer scientist. Those arc
things that I've done, but I've earned my
living as a computer programmer."

Plauger started working for Bell Labs
upon receiving his doctorate in 1969, first

By Craig LaGrow

at Holmdcl. N.J., then later at their Mur-
ray Hill, N.J., headquarters. He spent his
five and one-half years at Bell Labs as a
member of technical staff, which is the
standard staff level for people doing
research. His department was chartered
with getting computers into telephone
central offices and digitizing things so
that data processing technology could be
better brought to bear.

But he was also surrounded by people
like Dennis Ritchie, Ken Thompson, and
Brian Kernighan, who were working on
the beginnings of what was to become
UNIX and C. He naturally soaked up as
much of that technology as he could, fas-
cinated by anything that could improve
programmer productivity.

While Plauger was working at Bell
Labs, he designed small, in-house lan-
guages, built a few compilers, and began
to work with rudimentary UNIX and pro-
gram development tools.

"In many ways, the success of UNIX
was little more than a matter of Thompson
and Ritchie setting up fences, if you will,
to herd people in the right direction,"
remembers Plauger. "They just made it
easier for people to write reusable soft-
ware with good properties. If there has
ever been a central committee on how to
homogenize UNIX, it has only existed the
last couple of years — and a lot of the pay-
off has been lost in the growth and size of
UNIX from attempting to homogenize it."

According to Plauger, there is no ques-
tion that Ritchie invented the C language
and made it what it is today.

"It was his baby, and that's one of its
great strengths; its integrity stems from
the fact that there was one good designer
who was really responsible," says Plau-
ger. "All of us to this day who are
involved with it deferred to him."

klauger was one of
'the first real C
programmers.
Working on an old PDP-1 1/20— the verv
first PDP-1 1 UNIX system— he recalls'
tripping over many of the bugs when
things like structures first went into the
language.

An interesting footnote to the C lan-
guage development going on at the time
was that Plauger was simultaneously
developing a language he hoped would

rival C. He then completed a language
implementation that today would resem-
ble something like PL/M for the PDF- 1 1 .

"One of the smarter things 1 did in my
career was realize after a period of lime
that the C language was superior to
mine," he says. "I actually had the
decency to stand up and tell a lot of people
that even though I'd put a lot of work into
this language. C was better and they
should use it instead. I became a convert
in the early days."

Thompson had been experimenting
with a language called B. a precursor to C
that was, according (o Plauger, a "cute
derivative" of another experimental lan-
guage called BCPL. "The trouble with it
on a PDP-1 I, though, was that it assumed
everything was a 16-bit integer," Plauger
recalls.

The B language could manipulate point-
ers like 16-bit integers but, because of its
weakness with byte resolution addressing
of the PDP-1 I . a programmer had to
know to count by twos to step through an
array. Because of this, data typing went
into the language just to help the compiler
know how to manipulate pointers to dif-
ferent types of data.

The second problem with B was that it
wasn't a true compiler but rather an
interpreter.

"C in a nutshell is B with typing to get

SOFTWARE
DEVELOPERS!

V-FILE

THE VIRTUAL MEMORY

FILE MANAGER

Let V-FILE save precious development
time & cost as you create efficient appli-
cations with the power of VIRTUAL MEMORY.

DON'T RE-INVENT THE WHEEL

Why spend weeks or months coding and
debugging file and memoiy management
systems when you can order V-FILE today.
V-FILE is a library that you can link with
your code to provide sophisticated virtual
file and memory management — allowing
you to concentrate on developing your
application.

VIRTUAL DATA
OBJECTS SUPPORTED!

Data is referenced by using VIRTUAL
MEMORY DATA HANDLES. Your code
doesn't need to know whether the data
is actually on disk or in RAM. Swapping
between disk and RAM and updating files
on disk is handled automatically and trans-
parently! Complex VIRTUAL DATA
STRUCTURES can be created by linking
with data handles instead of pointers.

CHECK THESE FEATURES!

• Multiple, independent swap buffers

• Multiple files per swap buffer

• Highly efficient swap algorithm

• Automatic file updating

• Data prefetching supported

• Data may be locked in memoiy

• Memory buffers may be flushed

• Makes full use of extended memory on
IBM PC/AT

• SOURCE CODE AVAILABLE

• NO ROYALTIES REQUIRED

Supports Dos 2.00+ with

Lattice & Microsoft C compilers

Supports Microsoft windows

S299

^ Contact:
i pit p MindBank. Inc
* J^ 1 — 4620 Henry Street

Rttsburgh. PA 15213
^ TM 412/683-9800

VISA/MASTER CARD ACCEPTED

CIRCLE 63 ON READER SERVICE CARD
22 COMPUTER LANGUAGE! FEBRUARY 1985

the pointer arithmetic right and true com-
pilation to get better efficiency," says
Plauger.

Ironically, however, all the develop-
ment work on UNIX and C at the time was
being done in total opposition to the
administrative philosophy of Bell Labs.
The fact that the project survived, accord-
ing to Plauger, was mostly due to a policy
of "salutary neglect" on the part of
management.

"'There were no projects, no specs, no
requirements: it was almost all done on
speculation — all of it," says Plauger.
"Over about a two-and-onc-half-year
period, one by one most of the people in
our area at Bell Labs elected to do their
work on UNIX, but nobody told them to,"
he says.

The GE 635 computer, which was one
of the large systems that Bell researchers
were supposed to be using in their work,
actually turned into a remote printer for
the UNIX machine as far as this group
was concerned.

"Nobody made them do that. Nobody
made them write up the utilities which
eventually got swept up into the UNIX
package. Nobody enforced a uniform
style," says Plauger. "As a consequence,
we can all now piss and moan about the
irregularities that are there. But consid-
ering there was no enforcement, it's
remarkably regular. The miracle is that
UNIX is as good as it is. not that it
couldn't be better."

The early prototypes to UNIX and C
also were being used For many interesting
projects in-house at the time. Plauger
often reminds people that Thompson
started UNIX to support his own "little
solar system exploration thing" so it
would be easier to develop the software
for it.

"There were half a dozen people hang-
ing around the UNIX room because each
had his own little pet project going, and
UNIX was the best leverage to get that
accomplished," says Plauger. "Not a sin-
gle one of them was chartered by manage-
ment in order to improve productivity or
in order to pursue that particular goal."

"It was very easy to get caught up in all
this," he recalls. "Of course having Brian
Kernighan next door to shoot the bull with
and speculate about what we could do
with all these things was great."

l*^c

s Kernighan and
Plauger became
Lcloser friends
and work associates, they ended up writ-
ing two books on programming style
called The Elements of Programming
Style, ' Software Tools , 2 and Software
Tools in Pascal, 3 which have become text
book standards in many computer science
curricula across the country.

"I think Brian had a folder of examples
that showed how bad the training for pro-

grammers was at the lime," he says.
"Finally, we'd get to a 'there ought to be a
law' stage."

"What we did with Tlie Elements of
Programming Style was to write the manu-
script fairly quickly and circulate it
among our friends." he recalls, "We'd
tend to go on a kind of feeding frenzy,
tearing up programming examples from
other books, II took us a couple of iter-
ations to get to the correct style that I
think made the book what it is.

"Brian is a very solid writer, very
direct. And I've tried to emulate Issac
Asimov's science-fact writing," says
Plauger. "Between the two of us, Brian
and 1 managed to keep the worst offenses
of our programmer arrogance and love for
big words out of the document. It man-
aged to be something better than either
one of us could have done on our own."

Plauger remembers Ritchie being a bit
more distant to him both professionally
and socially at first. "He lived in his own
world, which was populated by the likes
of Ken Thompson and Doug Mcllroy,"
says Plauger.

"It was like, wow, this guy comes in at
1 :30or2:00 p.m. and immediately goes
to lunch! But nobody seemed to mind
because he'd work till midnight with these
people, and they'd all get tremendous
amounts of brilliant work done." he says.

Plauger admits that he'll always have a
tremendous respect and admiration for
Ritchie. "He's a marvelous writer and an
excellent designer with what amounts to a
slightly enhanced version of Appendix A
in Ritchie's and Kernighan's famous
book, Tlie C Programming Language .
Also of importance was the /usr/group
effort on standardizing libraries.

Serving as the secretary of the commit-
tee since its inception over a year ago,
Plauger was also appointed chairman of
the Library TG— one of the three tech-
nical subcommittees working on pieces of
the soon-to-be published ANSI X3J 1 1 C
standard.

"I think it's amusing that I ended up as
chairman of the Library TG because some
have said that's like putting the fox in
charge of the hen house,' ' jokes Plauger.
"I gravitated toward that because of the
tremendous technology that my own com-
piler company [Whitesmiths Ltd.) has
developed for making C work well— not
just in UNIX but everywhere."

"We all agreed to adopt the /usr/group
standard as the base document for the
library standards effort. Essentially what
I did was show the committee how to
modify that set of functions to get the
UNIX-isms out, so that they could be
implemented in a variety of
environments.

"We also have put forward and adopted
a number of additional functions which

you need outside of the UNIX environ-
ment. The net result is that what ANSI
now has is functionally very compatible
with UNIX V7 and System III, which
/usr/group has standardized on, but is
strongly enhanced semantically to make a
more widely usable standard," says
Ptauger.

Participating in the ANSI committee
meetings have been representatives from
over two dozen C compiler companies.
Consequently, the interests of many com-
panies comes to bear in each of the stan-
dards meetings.

According to Plauger, compiler manu-
facturing companies have a strong influ-
ence on the outcome of the standard-
ization. "Other groups, such as educators
and independent consumers, are often
outvoted because of this makeup," he
claims.

"Obviously one of the reasons that
we're having all these standardization
meetings is that we have these different
dialects that have evolved in different lit-
tle communities. What we're trying to do
is to bring them together."

Plauger feels strongly that the main
purpose of the ANSI standards effort is to
create a marketplace for the C language to
grow into.

"If there is no marketplace, there's no
way you can make money," says Plauger.
"You can bury your gold in your back
yard and be safe, but that's not a market.
Or at the other extreme, if there are no
rules for safe trade, there'll always be
pirates waiting out there every time you
send a ship out to do business. It'll get
boarded and everything gets ripped off.
That's not a marketplace either."

But today, after a year and one-half of
meetings, the committee is preparing to
finalize a document that will formalize the
rules and definitions that future C com-
piler writers will follow when designing
different implementations of C for differ-
ent hardware and operating system
environments.

"In these last few meetings, we're
really just getting down to the sorts of
things that Talmudic scholars would sit
around arguing about on a Saturday," says
Plauger. "We're really just quibbling over
whether certain nice extensions should get
stuck into the language before it's frozen
and what the subtle interpretation is of
certain types of coercions. Things like
that."

r et Plauger is also
willing to admit
that he thinks the
C language will soon begin to decline in
popularity within the programming com-
munity as a whole.

"My feeling is that C is just short of
peaking in popularity— that is, it has
almost peaked and will start declining
soon," says Plauger. "C has some mainte-

nance problems with large programs. It's
a little too lax in its declarations and its
type checking, and it doesn't have all the
facilities you need, such as for
packaging."

"You see the amount of stuff that's been
wriiten in C, from commercial applica-
tions to you name it," says Plauger.
"FORTRAN ranks in popularity right up
there with C, but C is unquestionably the
better language right now. Its major lim-
itation is that when you get a program-
ming project somewhere between 50 and
200 and 1 ,000 lines of code, you start
losing control."

"C just doesn't have all of the things
that you need for modularization, strong
checking, and soon. It's evolving in that
direction, but it's got its weaknesses.
Today certain dialects of Pascal are
becoming fairly powerful too. The trouble
with Pascal is that in order to meet those
requirements it has to be extended, and
nobody seems to have agreed on which
extensions yet.

"Ada is still out, of the starting gate. If I
had to do a large project commercially, I
would do it in structured COBOL. If I felt
the demand for performance were too
great. I might switch to PL/I but probably
to C," says Plauger.

Bell Labs is preparing to introduce an
enhanced version of C, called C+ + ,
which, according to Plauger, "has a lot of
bright ideas tacked on. But it's not com-
pelling to me that's the best thing to do to
C yet. Certainly a lot of good ideas that
are missing in C can be found in Ada. Ada
is full of a lot of things, it's just too big a
language. I think right now programming
in C, with tighter discipline, is what we're

going to be relying on to carry us through
the next few years," says Plauger.

ilauger's five-year
stint with Bell
Labs did not end
on the best of terms. In 1974, the woman
who he was married to at that time filed
charges of sex descrimination against Bell
Labs, and Plauger decided to support her
in her suit against the company because he
felt her claims were justified.

"I think that because 1 was so out-
spoken in encouraging her to oppose the
company made it hard for people to
want to keep me on at Bell Labs," he says.
"Let's face it, I didn't have the person-
ality to understand how to work well in an
organization of that size. I had a chance to
leave under my own steam before I got
run out of town on a rail, so I took it."

Plauger left Bell Labs in 1975 to
work for Yourdon Inc., a seminar and
book publishing company based in New
York, N.Y. Ed Yourdon, founder of the
company, had created a strong reputation
worldwide for being an author of
important books on programmer
productivity.

"I'd say I went to Yourdon with a head
full of ideas about programming tools,"
he remembers. "When the opportunity
came up to actually write a C compiler, I
was quite confident that I could do it even
though I hadn't written a whole C com-
piler before."

But Plauger soon left Yourdon to start
his own company on the kitchen table of
his apartment on New York's Manhattan

YOUR CODE MAY BE WASTING ITS TIME!
THE PROFILER™ CAN HELP . . .

• Statistical Execution Profiler • Time critical code optimization

• Works with any language • Abnormal code behavior tracking

• Completely configurable • Graphic presentation of results

• Up to 16 partitions in RAM/ROM • Easy to use menu interface

THE PROFILER is a software package which gives you, the programmer, a powerful tool for locating
time consuming functions in your code and allows you to performance tune your program. With
the THE PROFILER you can determine where to optimize your code for maximum benefit, then measure
the results of your efforts.

Using THE PROFILER, you can answer questions like:
Where is my program spending its time?
Why is my program so slow? Whal is it doing?
Is my progam I/O bound? CPU bound? Are data buffers large enough?
How much improvement did my changes make?
THE PROFILER is completely software based and consists of a system resident driver and a monitor
program. The memory partitions can range from 1 byte to 1 megabyte in size and can be anywhere
in the address space.

NO ADDITIONAL HARDWARE IS REQUIRED!
Requires an IBM PC or compatible system with a minimum 64k
and one drive.

THE PROFILER is available for $125.00 from DWB Associates or
ask your software dealer. To order or for more information, call
or write DWB Associates. VISA/MC accepted. Dealers welcome.

IBM is a trademark ol IBM Corp. MSD0S is a trademark Of Microsoft Corp
THE PROFILER is a trademark of DWB Associates.

€0

dwb

Aliaclati a

P.O. Bo* 5777
EJeavertoi\ Oregon 97006
(503) 629-9645

CIRCLE 20 ON READER SERVICE CARD

island. Ho called the company White-
smiths because it had a certain Old World
sound to it. "We wanted to call to mind
the old craft guilds almost back in the
Middle Ages, certainly the early Renais-
sance," says Plauger.

Even the company logo— which shows
a hand holding a ink quill pen in front of a
collection of different lettering styles
spanning thousands of years— can be seen
both as an ancient Roman seal and a cath-
ode ray tube.

"I have learned to think of program-
ming not so much as an art but as a craft,
meaning a more pragmatic art form."' he
says. "Certainly you don't write a pro-
gram and just have everyone stand around
and admire it."

Since that time, Plauger's company has
written C compilers that stretch from the
small, 8080-based microcomputers to the
IBM 370. Most of their customers, he
claims, depend on the fact that White-
smiths is able to bridge such a broad range
of systems and environments with a uni-
form language.

"Let's face it. 1 started this company
because I like writing code and showing
the world that you could really produce
high-quality code that is marketable. I
think we've done that," says Plauger.
"Now I'm learning how to be a
businessman.

"When you own your own company it

just tends to cat you up. " he says. "I do
pride myself on being a writer, but I've
had little time for that lately. And if you
ask me what I do for a living, I'd say I'm a
programmer."

| ot only is Plau-
ger the head of
his own

successful software company, but he's
also a talented science fiction author. He
won the John W. Campbell Award for best
new science fiction writer of the year in
1975. His story "Child of All Ages," 5
originally published in Analog , has been
widely anthologized and has been
optioned for television. His novel. Fight-
ing Madness,'' and several other short sto-
ries have been published by Analog and
other magazines.

Plauger emulates the prose style of
C.S. Forrester in his fiction, just as he
models his technical writing after
Asimov. He prefers to write in a clean and
straightforward way, without forcing his
readers to "bump their shins on the big
words or funky grammar." as he puts it.
"When you write science fiction, you
can't snow people with fancy terminology,
you have to sell a story about people."

Plauger now lives in a beautiful Early
Victorian house in Concord, Mass.. and
spends most of his life with his family
shuttling back and forth between this

home and another lakeside home in New
Hampshire. Above all, he says he enjoys
watching his five-year-old son, Geoffrey,
growing up.

Aside from programming and running a
company, Plauger practices piano daily
and manages to get in a jog every day as
well. "I'm a runner just forthe pleasure
of it, enough to keep my heart ticking well
and let me eat extra dessert."

As a multitalented individual, with
interests spanning from the artistic to the
pragmatic, Plauger certainly qualifies as
one of the more fascinating Renaissance
people in our computing industry today.

References

1. KernighLin. Brian W. and Plauger, P.J. The
Elements of Programming Style. McGraw-
Hill Book Co.. New York, N.Y., 1974.

2. Kcrnighan, Brian W. und Plauger, P.J. Soft-
ware Tools. Addison-Wcslcy Inc., Reading,
Mass., 1976.

3. Kcrnighan. Brian W. and Plauger, P.J. Soft-
ware Tools in Pascal. Add i son-Wesley Inc..
Reading. Mass.. 1981.

4. Kcrnighan, Brian W. and Ritchie. Dennis
M. Vie C Programming Language.
Prentice-Hall Inc., Englcwood Cliffs, N.J..
1978.

5. Plauger, P.J, "Child of All Ages." Analog,
(Nov. 1973).

6. Plauger, P.J. "Fighting Madness." Analog
Annual (1976).

Craig LaGrow is the editor of
COMPUTER LANGUAGE.

SMALL rFDRIBFFPG

Small-C CompilerVersion
2.1 for PC-DOS/MS-DOS
Source Code included
forCompiler& Library
New 8086 optimizations
Rich I/O & Standard Library

CBUG SOURCE LEVEL DEBUGGER FOR SMALL C

Break, Trace, and Change
variables all on the
source level
Source code included

Datalight

11557 8th Ave. N.E.
Seattle, Washington 98125
(206)367-1803

ASM or MASM is requited with compiler

i ncl ude d isk size < i 60k/3 20ki, and DOS versi on with order

visa & MasterCard accepted Include card no & expiration date

Washington state residents include 7 9% sales tax

IBM-PC & PC-DOS are trademarks of international Business Machines

MS-DOS is a trademark of Microsoft Corporation

CIRCLE 19 ON READER SERVICE CARD

24 COMPUTER LANGUAGE ■ FEBRUARY 1985

Only $95 with FULL SOURCE CODE!
"... an incredible learning tool," Byte

For only S95. Q/C is a ready-to-use C compiler for CPM with com-
plete source code. Heres what BYTE (May 19841 said. "Q-'C has
a portable library and produces good code quality. If you want to
learn compiler construction techniques or modify the standard lan-
guage. QC is the obvious choice

• Source code for compiler and over 75 library functions

• Strong support for assembly language and ROMs

• No license fees for object code.

• Z80 version takes advantage of Z80 instructions.

• QC is standard. Good portability to UNIX

Q'C has casts, typedef. sizeof. structure initialization, and function
typing. It is compatible with UNIX Version 7 C. but doesn I support
long integers, float, parameterized Adelines, or bit fields. Call about
our new products: Q C profiler. Z80 code optimizer, and Z80 as-
sembler and virtual linker, all with full source code!

miCODE

WORKS

5266 Hollisfer Suite 224
Santa Barbara. CA 93111
(805) 683-1585

QC CP M. Z80 ana UNIX art? trariemFsrks ol Oualiiy Compulei Svslems Digital
Research Zilog Inc and Bell Laboratories respectively

CIRCLE 42 ON READER SERVICE CARD

LOMAS DATA PRODUCTS INVITES YOU TO:

SHARE THE THUNDER.

t ii

The S100-PC -TM offers the
following standard features:

High performance THUNDER 186 8Mhz
80186 processor

512K bytes of RAM (expandable to 1Mbyte)
4 serial ports to support up to four users
3 Centronics compatible parallel ports
Concurrent DOS operating system
allows execution of both CP/M-86 and
MS-DOS (PC-DOS) programs
i . 5V4" IBM-PC compatible floppy drive

40 Mbyte high performance Winchester drive
Attractive 10 slot desktop enclosure

In addition, a number of options are available including:
larger Winchester drives, more user ports, 80286 proces-
sor, graphics support and additional operating systems
(MS-DOS and CP/M-86).

S100 BUS boards products & support
for the system integrator . . .

All of LDP boards are fully tested to exacting standards
and carry a one year warranty. We specialize in 16-bit
products & support the major operating systems for 16-bit
processors: CP/M-86*, CONCURRENT CP/M-86*, and
MS-DOS (PC-DOS).

■ THUNDER186 — THE ONLY COMPLETE S100 BUS,
16 BIT SINGLE BOARD COMPUTER AVAILABLE
TODAY.

Concurrent CP/M-86, which in addition to running CP/M-86
programs, runs MS-DOS programs. Comes complete,
ready to plug into an enclosure and run. 2S6K bytes of
RAM only PRICE $1595.00

■ LIGHTNING ONE***8086/8088 CPU

8086 or 8088, with 8087 and 8089 coprocessors. Up to 10
MHZ operation PRICES start at $425.00

■ HAZITALL SYSTEM SUPPORT BOARD

2 serial, 2 parallel ports, battery protected clock calendar.
Hard disk controller host interface PRICE $325.00

■ LDP 128/256K DYNAMIC RAM

Advanced dynamic RAM with LSI controller for failsafe
operation, parity PRICE 128K— $495.00, 256K— $795.00

■ RAM67 HIGH PERFORMANCE STATIC RAM

High speed (100ns) low power CMOS static RAM. 128K
bytes, extended addressing PRICE $995.00

■ LDP72 FLOPPY DISK CONTROLLER

Single/double density, single/double sided disks, both 8"
and SW" inch drives simultaneously PRICE $275.00

■ LIGHTNING 286— 80286 CPU BOARD

Offers 4 times the performance of a 5MHZ 8086 CPU while
maintaining software compatibility PRICE $1395.00

■ OCTAPORT 8 PORT SERIAL BOARD

to 19200 baud operation real time clock interrupt. Ideal
for multi-user systems such as MP/M-86* PRICE $395.00

S100-PC-TM is a trademark at Lomas Data Products, Inc.
*CP/M-86, MP/M-86 and CONCURRENT CP/M-86 are trademarks ot
Digital Research. **MS-DOS is trademark of Microsoft.
***Lightning One is trademark of Lomas Data Products, Inc.

LOMAS DATA PRODUCTS, INC.

66 Hopkinton road, Westboro, MA 01581
Tel: (617) 366-6434 □ Telex: 4996272

Dealer inquiries invited.

For orders outside the U.S., contact
our exclusive dealers: D Australia —
LAMRON PTY. LTD., (02) 85-6228
D Malaysia — EXA COMPUTER (M)

SENDIRIAN BERHAD, 795284

CIRCLE 54 ON READER SERVICE CARD

Standardization

By Jim Brodie, chairman of C Programming Language Standards

he C program-
ming language
has become
widely accepted
as one of the leading choices for a variety
of applications, including systems pro-
gramming (such as operating systems and
compilers), industrial real-time control
applications (such as telephone switching
systems and process control), and even
general applications (such as text editors,
data base systems, and spreadsheets).
This widespread interest and use in
many environments has prompted the
development of a standard to enhance the
portability of C programs by precisely
specifying the language, library, and envi-
ronment requirements of C.

In this article I will discuss a little of the
history behind this standardization effort,
the goals and working principles of the
standards committee, and how the lan-
guage currently being defined by the stan-
dard committee is different from the exist-
ing de facto standard for C. I will not
address the standard activities related to
the library or environment.

Why a new standard?

The C language has a de facto standard
definition in Tlie C Programming Lan-
guage by Brian Kernighan and Dennis

Ritchie (affectionately known as K&R in
C programming circles). The language as
described in K&R has been the starting
point for almost every C translator (com-
piler or interpreter) in use.

Despite the high quality of this book,
considerable divergence has occurred in
the C translator implementations where
language and library features or proper-
ties were ambiguously or imprecisely
defined (or simply not defined at all) . In
addition, the language has grown in vari-
ous directions as it has matured, and user
requirements and language weaknesses
have been identified. Unfortunately these
extensions to the C language have varied
significantly across C translator
implementations.

Because of the variations in the differ-
ent dialects of C, the portability of C
programs — one of the often noted
strengths of C— has suffered.

It became clear a need existed to define
a standard that would establish a precise
definition of what C translators should
accept and what C programmers could
count on when trying to write portable
programs. This led to the formation of the
C Programming Language Standards
Committee (labeled X3J1 1) under the aus-
pices of the American National Standards
Institute (ANSI).

The standards committee, which is
made up of volunteer representatives from
a wide cross section of industry, aca-

COMPUTER LANGUAGE ■ FEBRUARY 1985

demia, and C users, was formed in June
1983. The committee meets four times a
year in one week sessions to formulate the
new standard that will be presented to the
C community for comment and, if satis-
factory, approval.

The primary goal for the standards
committee is to codify existing practice. It
is not in the committee's charter to
develop a significantly different language
with many new features. Much of the
work of the committee centers around
determining what is existing practice and

trying to resolve the differences when
existing practice is inconsistent or
contradictory.

The committee feels strongly that the
resulting standard should enhance the por-
tability of C programs. The committee
uses the rule of "the fighting chance" — it
is striving to develop a standard that does
not force a C programmer to write por-
table programs (the low-level, machine-
dependent features of C are part of its
strength) but at least gives the program-
mer a fighting chance of writing a por-
table program if good programming prac-
tices are followed.

Maintaining the Spirit of C is also high
on the list of the committee's goals. Some
philosophical principles underlie the
original design of the C language. The
committee is striving to ensure that the C
language it defines stays very close to the
principles that made C successful in the
first place. Some of these principles could
be summarized in phrases like "Trust the
programmer and allow the programmer to
do what needs to be done," "Keep the lan-
guage small and simple, and "Make
it fast, even if it is not guaranteed
to be right."

This last principle may need a little
explanation. Many operations are defined
to be "how the target machine's hardware
does it" rather than by some general rule.
For example, the size of [he basic integer
variable type, im , is defined to be the nat-
ural (read fastest) integer size for the
machine. Another example is that C does
not define whether there is zero fill or
sign fill when an object is shifted to the
right'. It is allowed to be whatever is fast-
est and easiest on the particular target
machine.

The astute reader may have noticed that
the goals of allowing the writing of por-
table programs and permitting machine-
defined features arc frequently
diametrically opposed. A significant part
of the committee's work is determining,
on a case-by-casc basis, how the language
can be defined so that each of the goals is
reached in a reasonable way.

The committee is also working to
define a standard that will break as little
currently working code as is reasonably
possible. It has generally been accepted
that every translator will need to be mod-
ified to meet the requirements of the
emerging standard. However, it is the
committee's hope that only a small subset
rewritten to be accepted by a standard
conforming translator.

Other goals, including a desire for
semantic precision, internal consistency,
simplicity, and reliability, are also consid-
ered when trying to evaluate alternative
approaches and solutions before the com-
mittee. These goals often come into play
when considering possible extensions to
the C language.

Changes to K&R

The language portion of the standard is
being derived from Tlie C Reference Man-
ual by Dennis Ritchie, an early version of
which was published as Appendix A of

Vie C Programming Language . In the
remainder of this article I will outline
some of the differences between the cur-
rent draft of the standard and the language
as defined in K&R.

It should be noted that the draft is not
final and that the committee may change
its position on any of the features
described here. It is not possible to give a
final definitive list of language changes
and features since the committee remains
undecided on several fundamental and
many peripheral issues.

Despite this caveat, the features
described here will probably be in the
draft proposed standard submitted for
approval unless a significant amount of C
community feedback is received indi-
cating that the committee has made a
mistake.

Current practice is not limited to the
language as defined in K&R. Many
changes have become common across a
wide variety of C implementations.

Many of the current implementations of
C translators have established slightly
stricter rules for combining objects of dif-

funcK)

{

int addr=1234;

char * cp;

cp = addr; /* incorrect */

cp = (char *) addr; /* correct */

Listing 1.

struct firstst

(

int fl;
long si;

fcrent types in C expressions. This change
is also reflected in the standard. For
example, integers (with the exception of
0) can no longer be assigned to a pointer
variable unless an explicit cast
is supplied. Listing t shows an example of
how this requirement affects the writing
of code.

The standard also follows common, cur-
rent practice by extending the use of the
unsigned type modifier to char , short ,
and long variables (K&R only defined
unsigned int).

The requirement that names of mem-
bers of structures and unions be unique
across the entire program has been lifted.
Now each structure defines its own name
space, which means that member names
may be reused in different structures
without conflict.

Structures assignment as well as pass-
ing structures as arguments and having

struct simplst

{

int si;

short s2;

};

/* no name conflict with firstst. si */

struct simplst strl , str2;

struct simplst stf unc( ) ; /*declare a function returning a structure */

func( )

str2 = strl; /* structure assignment */

strl = stfunc(str2) ; /* function with a structure argument, */

/* returning a structure value */

Listing 2.

28 COMPUTER LANGUAGE ■ FEBRUARY 1985

functions which return structures are also
included in the standard. Listing 2 demon-
strates how code that uses these features
would look. Note that the assignment of
the structure variable strj to str2 causes
the contents of sir] to be copied into str2 .
This is not just a copying of pointers to the
structures. Future modifications to mem-
bers of strJ will have no effect on the con-
tents of str2 .

The type void has been included to
allow the specification that a function
does not return a value (i.e., is a pro-
cedure). This allows a programmer to be
more precise in specifying the correct use
of a function.

The listing below demonstrates several
uses of void. The declaration of the func-
tion yfitnc uses void to specify that this is a
function that returns no value. This is con-
trasted with the declarations of Ifunc ,
which returns a long int and ifunc which
returns an int (by default).

void vfuncQ;
long int IfuneQ;
ifunc {);
main()

(

vfuncQ;

(void) IfuncQ;

The main program in this example
invokes the void typed function vfimc as a
procedure that returns no value. The use
of void in a cast is shown in the last state-
ment in main . Here the void cast is used to
show that the long value returned by Ifunc
is being explicitly ignored after this call.
While the use of the cast in this situation is
not required by the standard, it does allow
for the writing of programs that are
clearer and more maintainable.

Simple enumerated types have also
been included. It should be noted that the
enumerated types of C are significantly
different than those supplied in some
other languages, notably Pascal. Enumer-
ated type variables are treated by C as
simple integer variables, and no prede-
cessor or successor functions are sup-
plied. It is also possible to control the val-
ues of each of elements in the enumeration
list. For example, in Listing 3 the enu-

meration constants apple , grape , pear ,
and orange have the values 0, 1,4, and 5,
respectively. The assignment causes
myjuice to have a value of 4.

The last line of Listing 3 emphasizes
the integer nature of enumerated type
variables. The variable myjuice is decre-
mented by one (using the operator).

Myjuice now contains the value 3 . not the
value I . No special processing is done to
limit the values held by an enumerated
type variable to those in its enumeration
list.

The rules for identifiers have been
changed slightly. Identifiers continue to
be of an arbitrary length: however, inter-
nal identifiers (e.g., auto and static vari-
ables) may now be significant anywhere
in the first 3 1 characters (as opposed to
eight in K&R). External identifiers (e.g.,
global variables and non-static functions)
are more restricted due to the large num-
ber of implementations that have assem-
blers, linkers, and other language-related
utilities that can not support long names.

Conforming implementations are only
required to support external identifiers
(global variables and non-static func-
tions), which arc single case and signifi-
cant in the first six characters. Imple-
mentation., can, of course, allow a greater
number of :>ignificant characters in an
external identifier.

The committee has also clarified the
rules in several areas where various
implementations had taken different
approaches. One area that falls into this
category is the "writability" of strings.
Some current implementations view
strings as constants that cannot be mod-
ified (and put into ROM, if useful). Oth-
ers view them as an array of characters

func2 ()

(

enum fruit (apple, grape, pear=4, orange};

enum fruit myjuice;

myjuice = pear;
— myjuice ;

Listing 3.

that can be modified just as any array can
be.

The current draft of the standard speci-
fies that strings are constants and are not
writable. It is allowable for a translator to
supply writable strings as an extension to
the language, however, portable programs
should not use this extension.

Considerable variety in imple-
mentations has occurred in the area of
macros. A fairly common practice in this
area is "pasting" together tokens by elid-
ing a comment between them. This prac-
tice, shown in Listing 4, is based on the
assumption that a comment can be
replaced by nothing rather than white
space as specified in K&R. This practice
will not have the desired merging effect in
standards conforming translators since
the standard explicitly stales that com-
ments are replaced by white space (e.g., a
blank).

Since the merging facility was viewed
as useful, the committee defined a new,
consistent syntax for token pasting on
Mefine lines. This syntax uses a new ##
operator (Listing 5).

New rules and extensions
In addition to the noted changes as a result
of common existing practice and clari-
fications, the committee has established
some new rules and added several exten-
sions to the C language to meet specific
well-defined needs.

The most ambitious extension has been
the addition of a facility, called function
prototypes, which allows the checking
and conversion of function call argu-
ments. This feature has been added to
address the C language weakness that
there was no way to ensure that actual
function arguments were correct other

Old style of merging

/* merge arguments a and b into a single token */
#define concat(a,b) a/**/b

Listing 4.

New style of merging

/* merge arguments a and b into a single token */
#define concat(a,b) a # b

Listing 5.

int protol (long, char *, short);
double proto2 (void) ;
unsigned int oldfunc();

funcQ

{

int abc;

char * ep;

short xyz;

double dbl;

abc = protol(abc, ep, xyz); /* abc is widened to a long */
dbl - proto2(abc); /* invalid function call */

Listing 6.

main()

{

unsigned char uc;

signed char sc;

char c;

int abc;

uc = '\377 f ;
sc = '\377';
c = '\377'

abc = uc;
abc = sc;

abc = c;

Listing 7.

/* assuming 8-bit char and two's complement/*

/* abc becomes 255 */

/* abc becomes -1 */

/* depends on the implementation, */

/* either 255 or -1 */

than by tedious hand-checking. Some of
the most difficult-to-find bugs are related
to mismatched types between a function's
parameters and the actual arguments on
the function call.

The optional prototype facility allows
the specification of a function declaration
that includes type information for each of
the parameters to the function. When
present, it causes the translator to check
each of the actual arguments on a function
call. If the arguments match the corre-
sponding parameter type in the prototype
then nothing is done.

If an actual argument has a different
type, but is convertible to the correct type
by the rules of assignment, then the code
to do the appropriate conversion is added.
For example, an int actual argument
would be converted to a long if the corre-
sponding argument in the prototype was a
long . If the arguments are different and
there is no valid assignment conversion
(e.g., between int and structure type) then
an error has occurred.

Listing 6 demonstrates the declaration
for several functions. Prow] is declared
to be a function that returns int . It has
three parameters which have types long ,
pointer to char, and short . respectively.
Protol is declared to be a function,
returning a double. It has no parameters
(note the use of the void keyword). Oid-
fiinc is defined as a function that returns
unsigned short . This declaration does not
contain any parameter type information,
and therefore the traditional rules that did
not include any checking or conversion
apply on the calls to oldfitnc .

The function calls made in June will be
handled as follows. Thecal! to protol will
be checked. It will be determined that abc
is of type ml and that the prototype calls
for a long . The appropriate conversion
code will be generated. The other argu-
ments match, therefore no additional spe-
cial work is required. The call \aproto2
will be checked, and since the prototype
says there are no arguments and the call
has an argument, the call is invalid.

There are several places in C where the
sign of a value is allowed to be whatever is
fastest for the compiler to generate. This
is seen when char and bit field values are
widened to integers during expression
evaluations. In these cases it is
implementation-defined whether a value

30 COMPUTER tANGUAGE ■ FEBRUARY 1985

[hat has the high-order bit set results in a
positive or negative integer value.

This freedom for implementors allows
for efficient code generation but causes
problems when the signedness of a char or
bit field value is important when writing
portable programs. This problem has
been addressed with Ihe addition of the
type modifier, signed, to the language.

The signed type modifier guarantees
that the translator will view the bit pattern
in the char variable or bit field as repre-
senting a signed value. Listing 7 shows an
example of the three types of signedness
that char variables can take on. Assuming
an 8-bit char, the assignment of " \377"
(octal 377) will set all of the bits in each of
the variables.

However, when these variables are used
(e.g., widened during the assignment to
an integer variable), they are interpreted
in different ways. The unsigned char is
guaranteed to take on only zero and posi-
tive values. The signed char includes both
positive and negative values. The "plain"
char has an implementation-defined
range. However, since implementations
are allowed to select the fastest widening
available in the hardware, plain char is an
appropriate choice when the signedness of
the value is not important (e.g., when it
will only hold printable characters).

Two new type modifiers have also been
added to the C language: const and vol-
atile . The const modifier has been added
to allow the specification that a data item
will not be changed during the course of
the program. This information is particu-
larly valuable in small environments
where constant data could be placed in
ROM, if desired.

The presence of the const modifier
causes the compiler to check for the
invalid direct assignment, incrementing,
or decrementing of the const object. Indi-
rect attempts to modify const objects via
pointers cause undefined behavior that the
compiler is not required to diagnose.

The code fragment in Listing 8 shows
how definitions using the const type mod-
ifier would look. The first line declares a
constant xyz with a value of 24. Next, the
constant array arr is defined and initial-
ized. Const objects should always be ini-

tialized since modifications are not
allowed in the body of the code.

The volatile type modifier indicates to
the compiler that this object may be mod-
ified in ways unknown to the translator. It
prevents the translator from optimizing
any expressions that use this object. This
has been added to address the need, in
many microprocessor environments, for
the adequate handling of memory mapped
I/O.

In a memory mapped I/O system, a read
or write of a location may cause special
side effects, even though the statement
appears pointless or unnecessary from the
C code point of view.

The code fragment that follows defines
and uses the volatile object iomem. If
iomem had not been volatile, the translator
would have been free to eliminate the first
assignment to iomem since there is no use
of iomem before it is assigned a new value
in the second statement. Since iomem is
volatile, this possible optimization is not
allowed. This preserves the special
semantics that may be associated with
repeated writes to memory mapped I/O
location.

main ()

(

volatile char iomem;
iomem = 0;
iomem = 1;

An important benefit of the volatile
modifier is that it frees the translator to do
sophisticated optimizations when it is not
present.

This previous discussion has been a
brief overview of the language-related
work which is currently going on in the
standards effort for the C programming
language. I have not presented all of the
changes and features that people will find
new and different in the standard. How-
ever, I have tried to cover some of the
major areas of change and to explain the
underlying goals and principles that have
guided all of the changes.

People interested in becoming involved
in the standards activity or in obtaining a
copy of the working document which,
when complete, will become the draft
proposed standard should contact either:

Jim Brodic

Motorola Microsystems
2900 South Diablo Way
Tempe, Ariz. 85282
(602)438-3456

or the committee vice chair:

Tom Plum
Plum Hall
1 Spruce Ave.
Cardiff, N.J. 08232
(609)927-3770 H

Jim Brodie has an M. S. in computer sci-
ence from Arizona State Univ. , Tempe,
Ariz. He currently works at Motorola in C
compiler development. He convened the C
Programming Language Standards Com-
mittee and is its chairman.

const short xyz = 24;

const int arr[4] = {188, 996, 44, -17);

func ()

{

int abc;

abc = xyz + arr[3]; /* appropriate use of constants
xyz =10; /* invalid use of a constant */
arr[2]++; /* invalid use of a constant */

Listing 8.

i We have over 200
complete, tested, and, documented functions.

All source code and demo programs are included.

The library was specifically designed for software
velopment on the IBM PC, XT, AT and compatibles. There are no royalties.

Over 95% of the source code is written in C. Experienced programmers

can easily "customize" functions. Novices can learn from the thorough comments.

louwmivi mnnuon^

RARY includes:

i Screen Handling Available Windows *> Full Set of Color Graphics
unctions Better String Handling Than Basic i DOS Directory and File Man-
gement Execute Programs, DOS Commands and Batch Files i Complete
ieyboard Control Extensive Time/Date Processing i Polled ASYNC
:ommunications General DOS/BIOS gate And More '

and DeSmet.

. rosoft, Computer Innovations, Mark Williams

C Compilers: Lattice C— $349, Computer Innovations C86— $329; Mark Williams C— $449.

C UTILITY LIBRARY $149

Order direct or through your dealer. Specify compiler when ordering. Add $4.00 shipping for
UPS ground, $7.00 for UPS 2-day service. NJ residents add 6% sales tax. Master Card, Visa,
check or P.O.

| ESSENTIAL SOFTWARE, INC

P.O. Box 1003 Maplewood, New Jersey 07040 914/762-6605

CIRCLE 27 ON READER SERVICE CARD

32 COMPUTER LANGUAGE ■ FEBRUARY 1985

C Instead

FORTRAN?

By Anthony Skjellu m

ne of the most
prevalent top-
ics for dis-
agreement
between programmers is the choice of
programming language. Many articles
discuss the relative merits and demerits of
various languages, and some authors
make invidious comparisons to show that
their own favorite language is belter than
the rest.

While many of these articles are quite
informative, they usually fail to bring out
how the application affects the choice of
language. For example, one would nor-
mal ly choose a data base language for data
base management applications as opposed
to a general-purpose language like BASIC
or Pascal.

Through examples this article attempts
to illustrate why the C programming lan-
guage works well for scientific and
engineering applications. Before begin-
ning, some background comments are in
order.

Traditionally, scientific and engineer-
ing applications have depended very heav-
ily on FORTRAN as the source language.
FORTRAN was introduced early in the
computer age. It subsequently became
widely available, and the language
afforded some portability between differ-
ent hardware. FORTRAN became popu-
lar with scientists and engineers.

For many of these programmers, the
concept of programming begins and ends
with FORTRAN. Unfortunately, FOR-

TRAN did not provide a programming
environment that was conducive to struc-
tured, modular programming.

Furthermore, the intense connection of
intrinsic functions to the compiler often
forced programmers to resort to complex
programming tricks. Many programmers
became experts on writing tricky pro-
grams instead of programs that could be
easily maintained, enhanced, or
understood.

In all fairness, FORTRAN has under-
gone some face-lifts to include structured
programming additions. But many pro-
grammers avoid these new features
because they are not supported uniformly
on different machines. Thus, traditional
FORTRAN 66 has remained the standard
form for programs.

New scientists/engineers automatically
learn FORTRAN because that is the lan-
guage they'll encounter in industry. A sort
of vicious circle exists. Part of the prob-
lem is that students taught FORTRAN in
an engineering/scientific setting often
don't learn to carefully distinguish their
FORTRAN code from an underlying
algorithm.

No computer language can force the
user into a perfect mold. The importance
of a structured computer language is that
it should make modular programming a
comfortable, natural effort. At the same
time, it shouldn't be so rigid that the pro-
grammer must work around language fea-
tures and restrictions.

C satisfies these requirements as a
structured language. It has the additional
advantage that it is widely available with
essentially no dialects. The standard
library defined for C varies somewhat
from implementation to implementation,

but a good level of software portability
exists.

We now turn to illustrations of the con-
venience and versatility of C in our appli-
cation area.

Learning to program in C

An important aspect of programming is
the language learning phase. C turns out
to be easy to learn, although advanced
features involving pointer arithmetic often
are not learned at first. Fortunately, quite
a bit can be done with just a few keywords
and knowledge of a minimum number of
library functions. Table 1 contains a list of
keywords and functions useful when first
learning C.

The relative ease of learning C is attrac-
tive for engineering and science students.
If necessary, rudiments of C program-
ming can be included as part of a second-
ary or college semester course, and stu-
dents can learn the language
incrementally as they develop programs.
(It is also worthwhile to require modular
program design and good documentation
as part of the programming task.)

If C programming is used with several
courses in a curriculum, students can
gradually build a considerable knowledge
of application programming in their field
without taking specific programming
courses.

In certain instances, students should
learn to deal with software/hardware
interfacing as part of course work. In
many circumstances, however, this is
avoided because students don't have time
to learn assembly language in an
engineering course.

However, since C handles address cal-
culations and binary logic with case, a
small assembly-level nucleus could be
prepared by the instructor, with students
developing the bulk of the code in C
instead of in assembly language. This
would allow students to get hands-on
experience they would otherwise miss.
Development would be much less painful
and time-consuming. This approach is
feasible in a majority of cases, and it can
often be used in practical applications
also.

With all these comments behind us.
let's turn to some example programs.
They are designed to convey some of the
flavorof C.

Example no. 1

Imagine the following problem. We need
to develop a bisection method algorithm
that can be applied to several distinct real
functions within a program. The bisection
method is a robust means for finding

zeros of functions. For example, if our
function were

f(x) = x- 1.0

then there would be exactly one zero,
namely whcn.v is equal to one. In more
interesting cases, we often can't deduce
the answer by inspection. (For more spe-
cifics on this method, consult Figure 1 .)

In such cases, we need a function to
find the desired zero or zeros. Such a pro-
gram is depicted in Listing 1 . (All listings
referred to in this article are available in
electronic form. You can download them
either from the COMPUTER LANGUAGE
Bulletin Board Service— (415)
957-9370— or take them from
COMPUTER LANGUAGE \s account on
CompuServe by typing "GOCLM".)

Our function _bisea() will find zeros of
the function called /(j . which we would
define elsewhere in our program. Some
valid examples for /(J are presented in
Listing 2. An example call to _bisectQ
might be the routine in Listing 3 .

As we defined our task, we'll want to
use a bisection method on more than one
function. For example, we might want to
perform bisection on three functions: /() ,
g() , and h() . To do so, we could write
three versions of the _bisectQ function.

This would work but is unnecessary and
also a repetition of effort. To avoid this,
we make the function under evaluation an
argument to a more general function
called biseclQ , which we present in List-
ing 4. A valid calling sequence to this new
routine might be as presented in Listing 5.

The importance of this improvement
should not be missed. Wc have gener-
alized the function to the point where il
can be applied like a mathematical oper-
ator. That is, the function to be worked on
is part of the argument list. Thus, bisect
operates on the function specified over the
interval selected.

C provides the ability to use functions
as arguments through the concept of
pointers to functions. The name of a func-
tion without parentheses is an address. So

Basic keywords and
functions in C

Keywords

Use

for

Used lor looping
structures

break

Break from looping
structures

continue

Continue ot start of
loop

if . . . else

Conditionals

return

Return a value from a
function

inl

Define integer type

float, double

Define real single,
double precision types

Functions

Use

exilfl

Unconditional
program termination

printffj

Formatted print
function

scanfj)

Formatted input
function

Description of the bisection method

Object

Find a solution of f(x) = (to within a tolerance) given a continuous function f|x) on the
closed interval [a,b]. The values f(a), f(b) must have opposite signs so that a zero exists in
the interval [a,b].

Algorithm

StepO:

Step 1 :

Step 2:

5tep3:

Step 4:

Table 1.

34 COMPUTER LANGUAGE ■ FEBRUARY 1985

Step 5:

Figurel.

See if f(a), f(b) have opposite signs if not, print error message and
abort program

Compute c = (a + b)/2

Compute t = f(c)

If 1 1 1 < tolerance, we are done. Return c as the location of zero.

Iff > Oandffb) >

b = c
If I > Oandf(b) <

a = c
IfKOandf(a) <

a = c
Iff < Oandf(a) >

b = c
This is the bisection of the interval

Go back to step 2

/* find a zero of f(x) */

/* limits */

/* convergence precision */

double _bisect(a,b,prec)
double a,b,"
double prec;

{

double fabs(); /* absolute value (standard function) */

double f(); /* this function returns double */

double curr; /* current point */

double temp_val; /* current value */

int cond; /* indicates condition of signs of problem */

if((f(a) <= 0.0)&&(f(b) >= 0.0))

cond =0; /* no reversal */
else
if((f(a) > 0.0)&&(f(b) < 0.0))

cond - 1; /* reversal*/
else

{

}

printf("Illegal bisection request [%e,%e]\n",a,b) ;
exit(l) ;

while(l)
(

/* loop 'forever' *•/

curr = (a+b)/2.0; /* bisect interval */
temp_val = f(curr); /* get value there *./

if (fabs(temp_val) < prec) /* we are done */
break;

if(temp_val < 0.0)

}

else

if ( !cond) a = curr;
else b = curr;

}

if ( !cond) b = curr;
else a = curr;

}7* end while(l) */
return(curr) ;

} /* end bisectQ */

Listing 1.

the word "cos" specifies the address of
the function cos() . The corresponding
argument variable fit in the function
bisectQ is defined as follows:

double (*fn}();

This means that_/)i is a pointer to a func-
tion which returns double precision real
numbers. Whenever we write (*fii)(x) , we
actually end up calling the originally
specified function, i.e., cos(x).

This provision is useful throughout pro-

gram development. It is one of the most
useful concepts of C. We can define very
general functions and pass other functions
to them to define their actual behavior in a
particular instance. Programmers not
only avoid repetitious coding but the

double f(x)
double x;

return(x - 1.0);

/* sin() function */
/* f(x) = sin(2x) */

double f(x)
double x;

{

double sin();
return(sin(2.0*x)) ;

}

double f(x)
double x;

(

double temp = 0.0;

double power = 1.0;

int i;

for(i = l;i < 6;i++) /* i loops from 1 to 5 inclusive */

temp = temp + power/((double)i)) ;
power = power * x;
}

return(temp); /* f(x) = polynomial */

Listing 2.

double y;

y"= J>isect(l. 0,2.0, .0001); h

range to search [1.0,2.0],
zero tolerance is .0001

Listing 3.

36 COMPUTER tANGUAGEB FEBRUARY 1985

double bisect(fn,a,b,prec) /* find zero of specified function */

double (*fn)(); /* function to find »zero of *■/

double a,b; /* limits */

double prec; /* convergence precision */

{

double fabsQ; /* absolute value (standard function) */

double curr; /* current point */

double temp_val; /* current value */

int cond; /* indicates condition of signs of problem */

if(((*fn)(a) <= 0.0)&&((*fn)(b) >= 0.0))

cond =0; /* no reversal */
else

if(((*fn)(a) > 0.0)&&((*fn)(b) < 0.0))

cond = 1; /* reversal*/
else

[

printf ("Illegal bisection request [%e,%e]\n",a,b) ;

exit(l);
}

while(l) /* loop 'forever' */

(

curr = (a+b)/2.0; /* bisect interval */
temp_val = (*fn)(curr); /* get value there */

if (fabs(terap_val) < prec) /* we are done */
break;

if(temp_val < 0.0)

{

if ( !cond) a = curr ;
else b = curr;

}
else

{

if ( !cond) b = curr;
else a = curr;

}/* end while(l) */

return(curr) ;

} /* end bisect() */
Listing A,

DeSmet
C

8086/8088

Development !>1f1Q

Package lUSJ

FULL DEVELOPMENT PACKAGE

Full K&R C Compiler

Assembler, Linker & Librarian

Full-Screen Editor

Execution Profiler

Complete STDIQ Library (>120 Func)

Automatic DOS l X/2.X SUPPORT

BOTH B087 AND
SOFTWARE FLOATING POINT

OUTSTANDING PERFORMANCE

■ First and Second in AUG '83 BYTE
benchmarks

SYMBOLIC DEBUGGER

s 50

Examine & change variables by

name using C expressions

Flip between debug and display

screen

Display C source during execution

Set multiple breakpoints by function

or line number

DOS LINK SUPPORT

s 35

Uses DOS .OBJ Format

LINKS with DOS ASM

Uses Lattice-' naming conventions

Check: □ Dev. Pkg (109)
□ Debugger (50)
D DOS Link SupL (35)

SHIP TO:

ZIP_

WAR

CORPORATION

P.O. BOXC

Sunnyvale, CA 94087

(408) 720-9696

All orders shipped UPS surface on IBM format disks.
Shipping included in price. California residents add
sales lax. Canada shipping add S5, elsewhere add
S15. Checks must be on US Bank and In US Dollars.
Call 9 am, - I p.m. to CHARGE by VISA/MC/AMEX.

greater generality often makes the func-
tions applicable for future purposes. Any
greater effort in carefully coding such a
function is rewarded by reduced effort on
a future project.

In introducing C. pointers to functions
should be discussed early. Students should
be taught how to use the device; expla-
nation of addresses and pointers can wait
until arrays have been introduced and
understood. In this context, the concept of
an address can be more easily motivated.

In this first example, we have illus-
trated an algorithm in its C incarnation,
and we also introduced the idea of passing
functions as arguments. When building
software for scientific/engineering appli-
cations, we are often faced with the need
to perform operator-like procedures on
functions we define. Often we have no
analytical representation for the func-
tions, so we must consider sonic sort of
interpolation. The next example illus-
trates how we can turn numerical data into
a "black box" and thus simplify other
computations.

Example no. 2

As a result of experimentation or numeri-
cal computation we often obtain a set of N
data points:

yDJ

1,2,

continuous. Once we do this, we can
potentially use functions such as bisecrQ ,
which accept a continuous function as an
argument.

The quality of answers that this tech-
nique produces depends heavily on the
method of interpolation we use. For cer-
tain applications, linear interpolation suf-
fices. Listing 6 is a general linear inter-
polation routine. This is the first step to
placing our data in a black box.

Given the general linear interpolation
routine, we can create a function that
appears to other routines as if it were con-
tinuous. If an .vis requested outside the
known interval, the interpolator will
detect this and return 1 .0e300, which is
nearly "infinity"* in double precision.

Error handling must be performed at
some level, but this is up to the program-
mer. Listing 7 provides the outer level
function, which uses linterpO on data
arrays called xarrayQ and yarraylj . These
arrays arc defined to contain npoint
points. We have been careful to make
linterpf) general so that it can deal with
many different data arrays instead of just
a specific pair. This makes life easy for
the programmer. For example, inverse
interpolation of the xarrayf} , yarrayfl ,
can be done without difficulty. To perform
this new calculation, we need only change
the line

y - linterp(xorray,yarrcjy,x,npoints);

of Listing 7 to the analogous form

x = linterp(yQrray,xarray,y,npoints);

From this data, we seek to obtain a con-
tinuous functional relation of the form
y =ffx) . Since we only know the relation
between .r , y at N points, we have to inter-
polate (or perhaps extrapolate) to get a
value we require. If we are successful, we
will be able to bury the discrete nature of
our data within a function that appears

double cos(); /* use cos() function for calculation */
double y; /* destination for answer */

y = bisect(cos,0.0,3. 14159, .0001);

bisect cos() function

interval [0.0,3.14159]

zero tolerance .0001
*/

CIRCLE 18 ON READER SERVICE CARD
38 COMPUTER tANGUAGEB FEBRUARY 19B5

Listing 5.

Linear Interpolation

double

linterp( x values , y values, x, points)

double

xvalues[]; /* source of x values (sorted) */

double

yvalues[]; /* source of y values (sorted) */

double

x; /* find y given this x */

int points; /* number of points */

int i; /* looping variable */

double low_x,high x; /* used for tabular x values */

double low_y ,high_y; /* used for tabular y values */
double y; /'* temporary for result */

if((x < xvaluesfO]) (x > xvalues[points-l ] ))

return(1.0e300); /* flag illegal result */

for(i = 0;i < points;i++)

{

low_x = xvaluesfi];

high_x = xvalues[i+l] ;

if((low x <= x) && (high x >= x))
r

low__y = yvaluesfi];

high_y = yvalues[i+l ] ;

break;
)
}

y = low__y + ((x-low_x)/(high_x-low_x))*(high_y-lowjj);
/* formula for a straight line */

}
/*

return(y);

Listing 6.

as seen in Listing 8. By inverse inter-
polation we mean that given a y value, we
can find the corresponding.v (possibly not
unique). Whether this is useful for spe-
cific data depends on the nature of the
data and the task at hand. Normally we
would have to re-sort the data before
attempting this so that y values became
ordered lowest to highest.

In this second example, we have shown
how to build routines to most con-
veniently structure discrete data within a

program. This approach avoids many
complications when using general rou-
tines and will come in handy when per-
forming many tasks in scientific/
engineering applications. Usually more
sophisticated interpolation will be
employed . These examples were included
to show the point, not to solve the general
problem.

C can be used successfully for the type
of applications considered in this article.
Since it is readily available, powerful, and

easy to learn, there is good reason to hope
that many engineers and scientists will
find C useful in future programming
tasks. H~

Anthony Skjellum graduated from Caltech
with a B.S. in physics in 1984. He is now
pursuing graduate studies in chemical
engineering, exploring the use of parallel
computers and distributed processing for
chemical engineering applications.

double f(x)
double x;
{

Listing 7.

/* black box form for data */
/* point to evaluate at */

double y;

extern double xarray[ ],yarray[ ]; /* our data defined

as external */
extern int npoints; /* points in arrays */

y = linterp(xarray,yarray,x, npoints) ;
/* compute answer */

return(y) ;

/* return answer (no error check) */

double g(y)
double y;

{

}

double x;

/* inverse black box form for data */
/* point to evaluate at */

extern double xarray[ ], yarrayf ] ; /* our data defined

as external */
extern int npoints; /■* points in arrays */

x = linterp(yarray, xarray,y, npoints) ;
/* compute answer */

return(x) ;

/* return answer (no error check) */

Listing 8.

40 COMPUTER tANGUAGEB FEBRUARY 1985

When the going gets tough, Optimizing C86
comes through time and time again, C86 is a
highly dependable C compiler that has been op-
timized through the years to provide the best com-
bination of reliability, speed, and performance,

FA5T. IN-LINE 8087/60287 SUPPORT

Now you can take full advantage of 8087/80287
capabilities, allowing your programs to run many
times faster than possible with other C compilers
Plus the source code to all routines is included,
so you have complete control over all functions.

• COMPATIBLE WITH WIDELY AVAILABLE
LIBRARIES such as HALO screen graphics
and many, many others {call for list).

• TOPVIEW SUPPORT LIBRARY provides win-
dowing capabilities.

• SPEED OPTIMIZATION — there's always room
to tighten your code, and Computer Innovations
has the tools to help. For example,
PROF1LER-86 helps identify key areas for
optimization.

TECHNICAL SUPPORT, NOBODY DOES IT
BETTER

Computer Innovations has earned a reputation
for providing customer support that is unequalled
in the industry. This includes a user's group, an
on-line bulletin board, and a user's newsletter.

MORE OF THE FEATURES YOU WANT

* SOURCE is provided to all libraries for totai
programming control. The source includes a set
of standard UNIX routines plus many DOS
specific functions.

• SPECIAL IBM-PC LIBRARY including com-
munication, screen, and keyboard handling
functions.

m COMPUTER

M INNOVATIONS, INC.

980 Shrewsbury Avenue Tinton Falls, NJ 07724

E 19Q4 Computer Innovations, \nc.

JOIN THE THOUSANDS OF PROGRAMMERS
WHO TRUST AND RELY ON C86

For Further Information Call 800-922-0169.

Technical Assistance Call (201) 542-5920.
Computer Innovations features a full line of
C products including C-to-dBase (dBase develop-
ment tool) and Introducing C (C Interpreter
Language Learning System). Call or write for a
product profile.

For Further Information Call

800-922-0169

Technical Assistance Call (201) 542-5920

CIRCLE 8 ON READER SERVICE CARD

An Open Letter
to
@ Users

Hi. I'm Leor Zolman. author of the BDS C Compiler and president of BD Software, inc. I would like
to lake this opportunity to personally thank all of you out there who have purchased and used BDS C
over the past five years for your overwhelming support, feedback, and especially your kind
words about the package to potential new users. As you may have noticed. I've been doing a lot of
advertising lately; while the advertising tells the world that BD Software, Inc. is still alive and
selling C compilers, nothing has served to popularize the package as much as word of mouth, and lor
that I am grateful.

While I'm on the subject of expressing appreciation, someone who really deserves a bunch of it is
Robert Ward, coordinator of the C User's Group. In the three-plus years that Robert has been in
charge of the Group, he has done a remarkable job of newsletter production and public domain
software acquisition. If any of you who use BDS C are not already enrolled as members of the C
User's Group, then you are missing out on some really xpeetmcmisr sofware available for
peanuts. Here are some of the offerings: a package that effectively makes BDS C put out assembly
language, screen editors, text formatters, a bulletin board system, cross-assemblers. Unix-style
libraries (if you don't like the standard CP/M-specific BDS C library), and lots more. Robert, you
done good.

BDS C first went on sale in August of 1979. For the first several years of its existence, all copies
were vended through one particular exclusive distributor in New York. Unfortunately, most dealers
and many users had trouble dealing with that distributor (to put it mildly), and many dealers
probably wrote off BDS C due to the difficulty inherent in obtaining it. Late in 1982 1 decided to
market BDS C myself, and therefore began to interact directly with a number of software dealers and
distributors. A number of those dealers and distributors have been a special pleasure to work with,
and I would like to list them here so that any authors out there considering the choice of
distributors might benefit from my experience: New Generative SyeteaeH. The Pregraaaier'B
Shop, Weetice. Werfcaaa aai Associates. In japan: Li f cheat lac. (not to be confused with
Lifeboat New Yorkl) In Australia: AICOM Pacific.

Thanks again to BDS C users, distributors and Robert Ward for a fun-filled, productive five year
adventure. If the next five years prove to be even half as gratifying as the last, I'm going to be one
lucky programmer...

wBn Wm©u

P.O. Box 2368

Cambridge, Ma. 02238

(61 7) 576-3828

42 COMPUTER IANGUAGEB FEBRUARY 1985 CIRCLE 3 ON READER SERVICE CARD

Parti — Parsing command line arguments

Programming
Macros

.art I of this art-
icle will cover
macros as build-
ing blocks for
parsing command line arguments in C
programs. Next month we'll explore how
macros can also be used as effective tools
when debugging large programs. Finally,
in Part III of this special series, we'll look
at how to use macros in C to translate a
common representation of data into dif-
ferent but related tables required by dif-
ferent programs.

Most programs take some arguments
from the command line. It is desirable to
have a standard way of parsing those
arguments to insure uniformity of the user
interface and reduce programming effort
spent on this common task.

By careful selection of macros, argu-
ment parsing can be specified in a very
compact manner that resembles data dec-
larations rather than executable code.
Most error conditions are handled in the
code generated by macro calls. The same
degree of compactness cannot be achieved
by using only functions without macros.

The parsing algorithm can be custom-
ized by setting values of several key vari-
ables. To reduce the volume of informa-
tion the programmer has to specify,
default values are provided whenever pos-
sible. The programmer can override each
default. A separate macro sets each key
variable. Each macro has at most one
argument and the name of the macro is
mnemonically related to the effect it
produces.

By Alexander B. Abacus

Command line arguments

Command line arguments are divided into
positional arguments and key-letter argu-
ments (also called options). Key-letter
arguments start with a trigger character,
which is usually a minus sign. The order
of processing positional and key-letter
arguments can be varied:

■ Deferred processing of positional argu-
ments. When all key-letter arguments are
processed before any of the positional
arguments, all key-letter arguments apply
to all positional arguments regardless of
the placement of key-letter arguments.

■ Immediate processing of positional
arguments. When all arguments are pro-
cessed in order, a key-letter argument is
effective only for the positional arguments
that follow after it on the command line.

■ Key-letter arguments must precede all
positional arguments— required by many
programs written in C. After detecting the
first positional argument, these programs
assume that all subsequent arguments are
positional. This approach is popular
because it is easy to implement. It is sup-
ported by the implementation described
here but is not recommended.

Key-letter arguments are further
divided into flag and text arguments. A
flag argument consists of a key letter fol-
lowing a trigger character. Several flag
key letters can be grouped together fol-
lowing one trigger character as in:

-ab
-a -b

A text argument consists of a single
key letter immediately following a trigger
character and followed by a character

string either immediately or after a white
space, as:

-t Word

-t "Quoted word"
-tTail

Handling of text arguments in C pro-
grams is inconsistent. Some require white
space after the flag letter, others do not
allow it, while most programs accept both
variations.

The input to the C program is the com-
mand line already parsed into words. A
word is a sequence of characters between
white spaces. White spaces can be part of
a word only if the word is enclosed in
quotes.

It is obvious that words passed to the C
program do not correspond to our key-
tetter arguments. In the first example
under flag arguments, there are two flag
arguments but only one word is passed to
the C program. In the first two examples
under text arguments, there is one text
argument but two words are passed to the
C program.

Solutions

The following macro definitions provide
tools for parsing positional and flag argu-
ments. Text arguments, additional error
checking, and some bells and whistles
will be added in the complete solution.
The macros defined in Listing I are
used as a "sandwich." Macros
ArgBEGINO , ArgLOOP, and ArgEND
should be thought of as bread in the sand-

wich. The meat between ArgBEGINQ and
ArgLOOP are the macros overriding pro-
vided defaults. Macros between
ArgLOOP and ArgEND define key-letter
arguments.

The complete solution is in the file
ARGPAR.H. which is far too long to be
printed alongside this article ( 1 pages in
length). If you would like a copy of this
listing, call the COMPUTER LANGUAGE
Bulletin Board Service (415 957-9370) or
pick it up on our CompuServe account by
typing "GO CLM" If you have no access
to a modem, send a stamped, self-
addressed envelope to the editor and he'll
mail you a copy. I strongly encourage you
to take this extra step if you're truly inter-
ested in further pursuing this subject.

The following macros were selected
from the file ARGPAR.H to illustrate the
more important and often fundamental
ideas on how to effectively use macros in
parsing command arguments in C.

ArgBEGlN(ArgumentCounter, Argu-
mentVector)

Macro ArgBEGINQ starts the sandwich
that defines argument parsing. Arguments
of this macro are typically the names of
the arguments of the function mainQ .

ArgTrigger(TriggerCharacter)

Any character can be specified as trigger
character. Default is ' — '.

ArgMin(LowLimit)

specifies the minimum number of posi-
tional arguments that must be coded on
the command line. If the macro is not
coded, the check is not performed.

ArgMax(HighLimit)

specifies the maximum number of posi-
tional arguments that may be coded on the
command line. If this macro is not coded,
the check is not performed.

ArgDescription(String Array)

The argument StringArray supplies the
description of command arguments. It is
printed when requested by the - .'option
or after any error message that terminates
the parsing of arguments.

ArgPosCall( Function)

This macro turns on the immediate pro-
cessing of positional arguments. The sup-
plied function is called when a positional
argument is recognized. At that moment
only those key-letter arguments that pre-
cede that positional argument have been
processed. (See section on function inter-
face.)

ArgKeyLeading

This macro forces the requirement that all
key-letter arguments must preceed all
positional arguments.

ArgLOOP

Macro ArgLOOP divides the
ArgBEGINQ —ArgEND sandwich in two
parts. The first part is optional and may be
empty. If present it sets the modes that
apply to argument parsing. All modes
have default values provided by
ArgBEGINQ . Each default can be over-
ridden separately by coding a macro
between ArgBEGINQ and ArgLOOP.
Those macros have already been
described.

The second part defines key-letter argu-
ments. Macros defining key-letter argu-
ments must be coded between macros
ArgLOOP and ArgEND . One (and only
one) such macro must be coded for each
key-letter argument. Macros defining
key-letter arguments are described next.

ArgFlagSet(KeyLetter, Counter-
Variable)

Character KeyLetter specified in this
macro can be used as a flag. This macro
increments the CoitnterVariable each time
that flag is found in command line
arguments. CounterVariable should be ini-
tialized to zero. Note that the flag is
allowed to occur more than once on the
command line.

ArgTextSet(KeyLetter, Pointer-
Variable)

Character KeyLetter specified in this
macro can be used to introduce text argu-
ment. Po'mlerVariable is set to point to the
first character of the text string. Pointer-
Variable should be initialized to NULL
pointer. Multiple occurrences of the text
argument specified by this macro are not
allowed.

ArgFlagCall(Key Letter, Flag Function)

Character KeyLetter specified in this
macro can be used as a flag. This macro
invokes FlagFunction . (See section on
function interface.)

ArgTextCallf Key Letter, TextFunction)

Character KeyLetter specified in this
macro can be used to introduce text argu-
ment. This macro invokes TextFunction .
(See section on function interface.)

ArgEND

Macro ArgEND ends the sandwich that
defines argument parsing. After the exe-
cution of the ArgBEGINQ — ArgEND
sandwich, ArgmnentCounier and Argu-
mentVector will be updated to reflect only
positional arguments.

Function interface

Functions required as the second argu-
ment of macros ArgFiagCallQ ,
ArgTextCallQ , and the only argument of
the macro ArgPosCallf) all have the same
interface. They should be defined as:

char * name(keyLetter, textString)
char keyLetter;
char * textString;

Character keyLetter contains the key
letter for key-letter arguments and ' \ 0'
for positional arguments.

Pointer textString points to the text of the
argument for positional arguments and
key-letter text arguments and is a NULL
pointer for key-letter flag arguments.

COMPUTER tANGUAGE ■ FEBRUARY 1985

Simplified argument parsing, positional and flag

#define ArgBEGIN(A_argc, A_argv) \

{ /* Begin block with local variables. Ends with ArgEND. */ \
struct \
[ int argc, *ptrArgCount ; \

char ** argVector; \

char * (*ptrFunction) (); \

char keyTrigger, *keyPointer; \

int arglndex, charlndex, nextPositional, keysTerminated; \
} aO; \

aO.ptrArgCount = &(A_argc); aO. argVector = (A_argv); \
aO.argc = *a0.ptrArgCount; aO. keyTrigger = '-'; \
aO.ptrFunction = argPos;

#define ArgPosCall(A_Function) aO.ptrFunction = (A_Function) ;
#define ArgTrigger(A__Trigger) aO. keyTrigger = (A_Trigger) ;

#define ArgLOOP \

aO. keysTerminated = 0; aO. nextPositional = 1; \

for ( (aO. arglndex = 1); (aO. arglndex < aO.argc); (++a0. arglndex) ) \

{ /* for all arguments */ \

if ( ( aO. argVector[aO. arglndex] [0] != aO. keyTrigger ) \
( aO. keysTerminated != ) \

( a0.argVector[a0.ar g Index][l] == '\0' ) ) \

/* not a key: bubble up positional arguments */ \
aO. argVector [aO. nextPositional] = a0.argVector[aO. arglndex] ; \
(void) (*a0.ptrFunction) ( '\0', aO. argVector [aO. nextPositional] ); \
++a0. nextPositional; \

} \- '

else /* it is a key */ \

{ \
for \

( (aO. charlndex = 1, aO.keyPointer = & aO. argVector[aO. arglndex] [1 ] ) \
; ( ( aO.keyPointer != (char *) ) \
&& ( (^aO.keyPointer) != '\0' ) \

) \
; (++a0. charlndex, ++a0.keyPointer) \

) \

{ /* for all characters in an argument */ \
switch ( a0.keyPointer[0] ) \

{

#define ArgFlagSet(A_KeyLetter , A_FlagCounter) \

case A_KeyLetter: ++A_FlagCounter ; break;

Listing 1. (Continued on following page) .

#define ArgFlagCall(A_KeyLetter, A_Function) \

case A_KeyLetter: (A_Function)( *a0.keyPointer , ); break;

#define ArgEND \

default: \

(void) fprintf \

( stderr \

,." Stop. Key letter \'%c\' rejected. \n" \

, *a0.keyPointer \

); \

exit (1); \
} /* switch */ \
) /* for all characters in argument */ \
] /* if key or not a key */ \
} /* for all arguments */ \

(* aO.ptrArgCount) = aO.nextPositional - 1; \
} /* End block begun with ArgBEGIN. */ \

static char * argPos( a_keyChar, a_argString )
char a_keyChar, *a_argString;
{ return ( (char *) ); }

#if defined DriverH

finclude <stdio.h>

void main(argCount, argVector)
int argCount;
char * argVector[];

[

int f_flag = 0;

ArgBEGIN (argCount, argVector)

ArgPosCall (argPos)

ArgTrigger ('-')

ArgLOOP

ArgFlagSet (*£', f_flag)

ArgFlagCall ('F', argPos)

ArgEND

exit (0);
} /* main */

#endif DriverH

Listing 1. (Continued from preceding page).
46 COMPUTERLANGUAGEBFEBRUARY198J

The returned pointer should be NULL
unless an error has been detected.

Predefined flags

The following key-letters are predefined
as flags:

— ? causes output of argument descrip-
tion. (See macro ArgDescriptionQ .)

-! enables all debugging code if the
program is compiled with debugging
code. This flag must be the first argument
after the command name. Besides acting
as a flag, it can also be immediately fol-
lowed by text, without a separating white
space. Text is used to enable selected parts
of debugging code.

The flag initializes the global variables
used by the debugging macros, to be
described in Part II.

A trigger character followed by

another trigger character signals the end
of key-letter arguments. Remaining argu-
ments are treated as positional even if they
start with a trigger character.

— If the trigger character stands alone
it is passed as a positional argument. This
is frequently used as notation for standard
input.

The test driver at the end of the include
file shows how to code the sandwich of
macros that have been described. Counter
variables referenced as arguments of
macro ArgFIagSelQ should be initialized
to zero. Pointer variables referenced as
arguments of macro ArgTextSetQ should
be initialized to a NULL pointer.

This is just the tip of the iceberg on how
macros can be used to improve C program
development efficiency. Next month we'll
focus on how to use macros in order to set
up a systematic method for debugging
your code before you're ready to
compile. gf

Reference

Harbison, Samua! P. and Steele, Guy L. C: A
Reference Manna!. Prentice-Hall. 1984,
pp.26-84.

Alexander Abacus has a B.S. in electrical
engineering and is a software consultant
for CCA Computer Inc. , Cranford, N.J.
His software experience includes work for
Sperry, IBM, and AT&T Bell Laboratories.

Debugging Bugging You?

Torpedo program crashes and debugging delays with
debugging dynamite for the IBM PC ... «

UP PERISCOPE!

First, you install the hardware.

The hardware's a special memory board
that fits in a PC expansion slot. Its 16K of
write-protected memory contains
Periscope's resident symbolic debugger. No
runaway program, however berserk it may
be, can touch this memory!

Then you UP PERISCOPE.

Use Periscope's push-button break-
out switch to interrupt a running
program ... even when the system's
hung! Periscope supports Assembly,
BASIC, C and Pascal. In addition to the
usual debugging capabilities, some of
Periscope's features are:

Stop your system in
its tracks at any time.

Use symbol names instead
of addresses.

Run a program on one monitor and
debug on another.

Monitor your program's execution

with Periscope's comprehensive

breakpoints.

Debug memory-resident programs.

Put your time to better use.

The Periscope system is $295. It carries a 30-day money-back
guarantee and includes the memory board, remote break-out
switch, debugger software, 100-page manual, and quick-
reference card. The memory board is warranted for one year. A
demonstration disk is $5.00.

System requirements for Periscope are an IBM PC, XT, AT or
Compaq, PC-DOS, 64K RAM, 1 disk drive and an 80-column
monitor. For MasterCard and Visa orders only, call 800/421-
5300 (ext. R%) 24 hours a day. For additional information, call
404/256-3860 from 9 AM to 5 PM Eastern Time.

Get your programs up and running;

up PERISGl^E!

Data Base Decisions / 14 Bonnie Lane /Atlanta, GA 30328
CIRCLE 30 ON READER SERVICE CARD

POWER TOOLS

You are a serious programmer. Your income depends on developing
software solutions for complex problems. Fast. Error free. You need
powerful tools for your compiler. That's what we sell.

PLINK ™ 86

OVERLAY LINKAGE EDITOR

overlays for modules are automatically loaded during ex-
ecution

generates symbol tables for PFIX 86 PLUS
accepts unlimited number of files

capacity for over 35,000 symbol, commonblock and vari-
able names

change overlay structure without recompiling

sorted memory map reports

user defined width/height for reports

define symbols as absolute addresses or offsets from other

symbols

overlay is user configurable

List: S395

PFIX ™ 86 PLUS

DYNAMIC SYMBOLIC DEBUGGER

debugs PLINK 86 overlays

view source code while debugging

set temporary or permanent breakpoints

advanced tracing and trapping capability

in-line assembler for temporary patches

preserves program screen and supports dual CRT con-
figurations

multiple windows to view code, data, breakpoints, regis-
ter/stack contents, simultaneously
supports 8087 numeric coprocessor
List: S395

PMATE ™

TEXT EDITOR PROCESSOR

automatic disk buffering

10 auxiliary buffers that can be edited

garbage stack

user customization

120 command macro language

store macros permanently

arithmetic and logical operations

100 numeric variables

automatic formatting with indent and word wrap

insert, overwrite and command modes

horizontal/vertical scrolling

List: $225

[PC]

We Sell Power

Phoenix Computer Products Corporation

1416 Providence Hwy., Suite 220

Norwood, MA 02062

800-344-7200 (Mass.) 617-762-5030

Visa • Mastercard • American Express

COMPUTER LANGUAGE ■ FEBRUARY 1985

CIRCLE 82 ON READER SERVICE CARD

to Assembly
Interface

|he fundamental
question that
arises when con-
fronted wilh a
new general purpose programming lan-
guage is '"why use it?", which really
means "what advantages docs it have over
other languages?"

Of course, the answer depends upon the
type of programming tasks to be carried
out and their complexity or magnitude.
For complex programming tasks, any
advantages will have considerable bearing
on both the effort required to accomplish
the task and the quality of the resultant
program.

It is necessary to access architcclurc-
dependent machine capabilities to effi-
ciently accomplish most complex pro-
gramming tasks. Even without
considering efficiency this is readily
apparent for systems software. Since it is
the nature of any non-machine language to
be independent of these architecturally
dependent features, a mechanism is nec-
essary to access these features. The rela-
tive ease and efficiency of reaching these
features is a consideration in choosing a
language.

In this article we will briefly look at the
advantages of C over other compiled pro-
gramming languages and then focus on
how C can be easily interfaced to access
architecture-dependent capabilities. In
particular, we will show how Lattice C
can be efficiently interfaced to IBM
MACRO Assembler to access the IBM PC

By Brian H. Burger

ROM interrupt system by providing
source code to perform video I/O. The
code provided can be extended to access
other architecture-dependent features on
the IBM PC or on any computer using an
8088/86 architecture.

High-level vs. assembler

But if one must access these architecture-
dependent features in any case, why not
write the entire program in the assembler
language of the target computer? After
all, assembler code should result in the
most efficient object code.

The problem is that assembler code is
Ihc most time-consuming to write and
understand after it is written and will
result in a large amount of source code to
monitor and maintain. In addition, the
program will only work on the target
computer (or processor) for which it is
designed.

It is generally far superior to write pro-
grams in a high-level language for
increased productivity and portability.
Thus, the better high-level language is the
one that is most productive and most por-
table since that is the reason for using it in
the first place.

Yet the resultant program must still use
computer resources efficiently. There is
no advantage to using a language that pro-
duces a program too slow or too memory
hungry, even if it can be generated quickly
and will work on lots of machines—
except perhaps when that program is
used, not as the end result, but as part of
the process of development.

A superior language generates a more
resource-efficient program. In fact, dur-
ing the early development of FORTRAN
the efficiency of the program resulting

from the language was considered to be of
much greater importance than the design
of the language itself. John Backus wrote
in The History of Programming
Languages : '

To this day I believe that our emphasis
on object program efficiency rather than
on language design was basically correct.
I believe that had we failed to produce effi-
cient programs, the widespread use of lan-
guages like FORTRAN would have been
seriously delayed. In fact, I believe that we
are in a similar, but unrecognized, situ-
ation today (1978): in spite of all the fuss
that has been made over myriad language
details, current conventional languages
are still very weak programming aids, and
far more powerful languages would be in
use today if anyone had found a way to
make them run with adequate efficiency.

Separately compiled subprograms

A high-level language is used instead of
assembler when the benefits of increased
productivity and portablity outweigh the
penalty of less efficient use of computer
resources. So the best high-level language
to use is the one which is most productive,
most portable, and generates the most
efficient object code.

Another feature of a superior language
is its case of use and, by implication, pro-
grammer productivity. A language is eas-
ier to use if one can separately compile
any number of subprograms and link them
together later to produce an executable
program. A language that allows sepa-
rately compiled subprograms is superior

T ENDS

DSPIO

PROC

NEAR

POP

INT

10H

CMP

SP.BP

DONE

SAVE RETURN ADDRESS
GET PARAMETERS

; OPTIONAL POINTER TO 4 WORD AREA

; BP IS ALWAYS >= OFFSET OF LAST ARGUMENT
; IF NO RETURN POINTER PASTITLE DSPIO -
; SUBROUTINE TO PERFORM DIRECT SCREEN I/O

int ax, bx, ex, dx ;
char retval [8] ;

dspio( ax, bx, ex, dx, retval ) ;
dspio( ax ) ;
dspio( ax, bx ) ;
dspio( ax, bx, ex ) ;
dspio( ax, bx, ex, dx ) ;

ax - dx VALUES TO BE PLACED IN THE REGISTERS OF THE SAME NAME
AS NEEDED BY PARTICULAR VIDEO I/O FUNCTION

retval, IF SUPPLIED IS A POINTER TO A 4 WORD AREA INTO WHICH THE
4 REGISTERS AX - DX ARE COPIED AFTER RETURN FROM INT 10H.

DGROUP GROUP DATA

DATA SEGMENT WORD PUBLIC 'DATA*

ASSUME DS: DGROUP
DATA ENDS

PGROUP GROUP PROG

PROG SEGMENT BYTE PUBLIC 'PROG'

ASSUME CS: PGROUP

PUBLIC DSPIO

RLAYOUT STRUC
RAX DW ?
RBX DW ?
RCX DW ?
RDX DW ?
RLAYOUT ENDS

Return address space layout; pointed at by retval

DSPIO PROC NEAR

POP SI ;

POP AX ;

POP BX

POP CX

POP DI ;

INT 10H

CMP SP,BP ;

JA DONE ;

MOV [DI].RAX,AX

MOV [DI].RBX,BX

SAVE RETURN ADDRESS
GET PARAMETERS

OPTIONAL POINTER TO 4 WORD AREA

BP IS ALWAYS >= OFFSET OF LAST ARGUMENT

IF NO RETURN POINTER PASSED, DON'T SAVE VALUES

Listing 1 (Continued on a following page).
50 COMPUTER LANGUAGE ■ FEBRUARY 1985

lo one that docs not. It wastes computer
time to recompile an entire program
whenever a change is made, and it wastes
programmer time to have to wait for a
compilation before proceeding.

Without the capability for separate
compilation of subprograms, it is
extremely difficult to maintain a library of
useful subprograms, next to impossible
for more than one programmer to work on
the program, and difficult— if not
impossible— to access architecture-
dependent capabilities. Certainly, useful
programming techniques like information
hiding, stubbing, modular design and
development, and others are compatible
with separate compilation of subprograms
and are at least inconsistent with the idea
of one monolithic program compiled as
one unit.

The language that produces the most effi-
cient object code, is most portable, allows
for the separate compilation of sub-
programs, and is most productive to use is
the best language. C is recognized as pro-
ducing efficient object code because it
deals with the same sort of objects that
most computers do. Constructs like
register-type variables, painters, the
increment/decrement operators (+ + ,

, + = , * = , etc.), bitwise logical

operators ( < < . > > , &, etc.). a macro
capability, and others all lead to efficient
use of machine resources.

C is considered to be a very portable
language. ; It allows and in fact encour-
ages separately compiled subprograms.
Part of its portability power and efficiency
is a consequence of not providing lan-
guage constructs to do architecture-
dependent machine operations but instead
to encou rage the creation of standard
libraries to accomplish these tasks.

The brevity and elegance of its expres-
sive power leads to increased productiv-
ity. Note that I say "leads to." C requires
an understanding of the elements of the
computer and related concepts to a greater
extent than many other high-level pro-
gramming languages. The average

COBOL programmer has little use for
knowledge concerning the difficulties
encountered when a data structure falls on
an odd or even byte address boundary
since COBOL assumes the programmer
isn't concerned with these situations and
consequentially accepts any resultant
inefficiencies.

Since C doesn't provide any language
constructs to accomplish architecture-
dependent machine functions, it is partic-
ularly important that one can interface C
efficiently to assembler to accomplish
these tasks. Even though C compilers nor-
mally include a standard library of sub-
programs, some machine-dependent fea-
tures are not always included. It is not
practical to expect a language or compiler
system to provide access to all or even
most of these features since they arc spe-
cific to a particular computer or architec-
ture and frequently specific even to a par-
ticular configuration.

Architecture-dependent capability

But what is meant by accessing
architecture-dependent machine capabili-
ties? For example, machines frequently
provide in ROM a set of routines for per-
forming machine specific operations. On
the IBM PC and other machines utilizing
the Intel 8088/86 processors, these rou-
tines can be accessed in assembly lan-
guage through the software initiated inter-
rupt system.

An interface to an assembly language
subprogram is therefore needed to access
them from a high-level language. In
another example, machines generally
have multiple processors that work inde-
pendently of the main processor from
which they get instructions. In 8088/86
assembly, an instruction can be sent to one
of these processors by using an assembly-
level instruction named OOTdesigned for
this purpose.

The basic problem then becomes inter-
facing from a high-level language to the
assembly level to access these special fea-
tures in an efficient and uncomplicated
manner. The rest of this article is devoted
to doing just that by illustrating the ease
with which C can be interfaced to assem-
bly to access the IBM PC video interrupt
routines.

You may be wondering why we aren't
interfacing to perform DOS calls instead.

in particular the DOS version 2 extended
screen and keyboard device driver. While
doing this will likely make the program
more portable for DOS machines, this
DOS driver is slow and cumbersome. In
any case, similar techniques can be used
to interface to DOS interrupts.

Interfacing C

Specifically, Lattice/Microsoft C
(referred to from this point forward as just
C) is interfaced to IBM MACRO Assem-
bler (MASM) to perform video I/O as
shown in the MASM source code in List-
ing 1 and the C source code in Listing 2. 3,4
Figure 1 shows the C memory model that
is being used. Its limitations are that the
program can't be larger than I28K— the
program segment is limited to 64K, and
the combined data and stack segment is
limited to 64K,

This discussion will not be concerned
with C programs that use an extended
memory model. If a program needs an
instruction segment or data segment that
exceeds 64K bytes, unavoidable ineffi-
ciencies creep into the implementation.
There are many other ways to address this
need rather than extending the memory
model of the compiler. A discussion of
this subject is beyond the scope of the
present article. Furthermore, this model
is insufficient in few cases.

Upon entry into a subprogram from C,
the BP is the only general purpose register
that needs lo remain unmodified or be
saved. C loads the BP with a pointer to the
top of the stack before any arguments are
pushed and uses that value to deallocate
the stack upon return from the sub-
program. This feature allows a variable
number of arguments to be passed to the
called subprogram, as the caller can deal-
locate the arguments from the stack by
simply restoring the BP to the SP after the
call. This mechanism requires a two-byte
instruction (MOVSP.BP) to restore the
stack after the call instead of the three- or
four-byte instruction required to explic-
itly add a value to the stack pointer.

Since it is the caller's responsibility to
deallocate arguments from the stack,
there is no need to know at compile time
the number of arguments to be passed. AH
other registers can be freely used except
for the segment registers which must
remain unchanged from their original

C Is The Language.
Lifeboat Is The Source.

Lifeboat.™
The Leading Source And Authority For Serious Software.

1-800-847-7078.

In NY State: 212-860-0300

Serums Software For The C Programmer From Lifeboat.™

Lattice" C Compiler: The serious software developer's first choice.

Selected for use by IBM,'^ Texas Instruments, Wang," MicroPro," Ashton-Tate,"
IUS/Sorcim," Microsoft" and Lotus" to name a few of the many. Why?
Lattice C is clearly the finest 16 bit C compiler available today.

— Renowned for speed and code quality.

— Fully compatible with the C standards set forth by Kernighan and Ritchie.

— Four memory model options offer you unsurpassed control and versatility.

— Superior quality documentation.

— Now includes automatic sensing and use of the 8087 chip.

— Widest selection of supporting add-on packages.

Halo": A graphics development package rapidly emerging as the industry standard.

— Over 150 graphics commands including line, arc, box, circle and ellipse primitives.
— New: multiple typefaces which may be scaled and rotated.
— Supports multiple viewports and is device independent.

C Food Smorgasbord'": This beautifully written collection of C functions is a valuable time saver.
— Library includes a binary coded decimal arithmetic package, level I/O functions,

a terminal independence package, IBM PC ROM BIOS access functions and much more.

Pmate": The premier editor for the programming professional.

Pmate is a full screen editor with its own powerful macro command language:
— Perform on screen row and column arithmetic, alphabetize lists,

translate code from one language to another, call up other macros.
— Customize Pmate almost any way you like.

— Contains 10 auxiliary buffers for storage of macros, text, subroutines.
— An "undo" feature allows the programmer to retrieve whole series of deleted items.

Additional C Tools Panel™: Screen formatter and data entry aid.

Available From Lifeboat: Lattice Windows™: Windowing utility; create "Virtual Screens."

Plink-86 ,M : The popular linker; includes extensive overlay capabilities.

Pfix86' M : Dynamic debugging utility.

Pfix86 Plus'": Symbolic debugger with capacity to debug overlays.

Btrieve": Database record access/retrieval library.

Phact: Multikeyed ISAM C-Function library.

Fabs: Fast access B-tree database function library.

Autosort: Fast sort/merge utility.

ES/P: 'C program entry with automatic syntax checking and formatting.

Greenleaf Functions'": Library of over 200 popular C functions.
And much more.

YES! Please rush me the latest FREE Lifeboats catalog ofC products.

' Company Business

. Name Phone

Name Title

A ddress

City State Zip_

Please check the category where Lifeboat can best help you:
U Software development \ 3 Corporate Q Education
D Dealer! distributor Q Government □ Other

IMP Fli "[-w.ri, n

Call Direct: 1-800-847-7078 (In NY Stale: 212-860-0300)

Return coupon to: Lifeboat Associates™

1651 Third Avenue, New York, NY 10128.

CIRCLE 64 ON READER SERVICE CARD

MOV

[DI].RCX,CX

MOV

[DI].RDX,DX

JMP

DONE:

PUSH

DX ;

PUSH

CX ;

PUSH
JMP SI

DSPIO

ENDP

PROG

ENDS

END

RESTORE STACK BECAUSE OF THE POSSIBILITY THAT THEY
WERE NOT PASSED AS ARGUMENTS

Listing 1 (Continued from a preceding pane).

"41

main( )

{
int retvals
int index ;

SET_MODE ;

hline(0, 0,219,80) ;

hline(24, 0,219, 80) ;

verline(0, 0,219,24)

verline(0,79,219,24)

window(3,3,6, 5) ;

window(10, 10,5,6) ;

window(7,17,10,ll)

window(15,40,10,6)

window(5,50,20,16)

/* r ead c#define PAGE
/* video modes */
# define s40x25_bw
fdefine s40x25_color 1
#define s80x25_bw 2
#define s80x25_color 3
#define med_color 4
#define med_bw 5
#define high_bw 6

/* video functions */
#define set_type 256
^define set_cur 512
#define read_position 768
#define read_light_pen_position 1024
#define select_page 1280
#define scroll_up 1536
#define scroll_dn 1792
#define read_attribute__char 2048
#define write__attribute_char 2304
#define write_char 2560
#define set_palette 2816
#define write_dot 3072
#define read_dot 3328
#define write__teletype 3584
#define get_state 3840

/* draw box around screen */

/* i.e., manual window using nonwindow characters */

macros

#define curpos(row,col) dspio(set___cur ,PAGE,0, (((int)row)<<8)+col)

Listing 2 (Continued on a following page).
54 COMPUTER LANGUAGE ■ FEBRUARY1985

entry values. This mechanism is efficient
and flexible— only one register needs to
be saved upon call of a subprogram or
remain unmodified, and it can be passed a
varying number of arguments.

C passes arguments to its subprograms
by value, not by reference. So any argu-
ment passed is a copy of the original
value, not a reference or pointer to it.
Arguments are passed by pushing them
onto the stack in reverse order. That is,
the first argument pushed is the rightmost
argument appearing in the C function call.
Upon entry to the subprogram, the first
parameter appears on the top of the stack
after the return instruction pointer pushed
by the call.

The routine

The subprogram in Listing 1 is straight-
forward. To understand the particulars of
the video I/O functions sec the IBM Per-
sonal Computer Technical Reference Man-
ual. The macros in the source code in
Listing 2 provide enough information to
use most of the functions without the man-
ual, and the source code preamble in List-
ing 3 provides most of the additional par-
ticulars needed.

The video I/O interrupt lakes its argu-
ments from the four general purpo.se reg-
isters AX through DX and returns values
in those same registers. This subprogram
takes five arguments corresponding to the
four general purpose registers to be
passed to the video I/O interrupt routine
and a pointer argument used for storage of
any return values. The return address is
popped into SI, the arguments are popped
off the stack directly into the four general
purpose registers, and the fifth argument
is popped into the DI register for later
use. The second through fifth arguments
are valid only if the subprogram was actu-
ally called with those arguments.

After the video I/O interrupt is exe-
cuted, the subprogram determines if a
pointer was actually passed a.s (he fifth
argument by recognizing that the BP is
equal to the SP before any arguments
were pushed. If it is a valid argument, the
values of the four general purpose regis-
ters are copied to the locations pointed at
by the fifth argument.

Similarly, all but the first argument are
returned to the stack in case they were not
actually passed. For those functions that

return values in the general registers in
addition to AX, all five arguments must
be supplied. Otherwise incorrect values
would get on the stack, and return values
would be placed somewhere other than
where they were supposed to go.

Only those arguments actually needed
must be passed if no values are returned,
plus any intervening arguments. When
dspio is called to perform the set mode
function in Listing 2, only one value is
passed. Since the video I/O interrupt
doesn't care what is in any of the registers
it doesn't use when called, there is no
need to pass trailing filler arguments.

However, if the re are some intervening
arguments, as when changing the cursor
position in curpos , the intervening argu-
ments must be included. To avoid the
passing of filler arguments, the function
code passed in the first argument can be
used to branch to code that pops the argu-
ments only into registers pertinent to the
function. Thus, for function code 2 (i.e.,
set mode), it can be assumed that there
will be no argument supplied for CX and
the third argument can be popped into DX
instead.

This is what dispio in Listing 3 is all

C memory model

about. Dispio avoids passing arguments
unncccessarily, thereby increasing exe-
cution speed but for the penalty of
increased instruction overhead.

An alternative

In Listing 3 a jump table is used to effi-
ciently control the execution of code. It
works by storing offsets in a table in such
a way so that the function code stored in
the first argument can be used as an index
into the table to obtain the appropriate
address to jump to. Thus, the order in
which labels are stored into CASEJTABLE
determines the code that will be executed
upon receiving a function code in the high
byte of the first argument. The first argu-
ment is copied into BX . the Tunction code
is extracted, multiplied by 2 (since
addresses are words), and used as an
index into CASEJTABLE to obtain the
address to which the subprogram trans-
fers control .

C treats subprograms a.s functions that
return values. They return their function
value to the caller in the general registers.
The registers pertinent are determined by
the type of the function. Thus, for/«r

At most 64K segment
of data and stack

At most 64K segment
of instructions

Grows

downward

Grows

upward

Grows

upward

Stack segment

Data segment

Control segment

Figure 1.

#define SET_MODE dspio(s80x25_bw)

#define horline(ch, width) dspio(write_char+ch, PAGE, width)

#define hline(row, col, ch, width) curpos(row,col) ; horline(ch, width)

#define wrtchar(ch) dspio(write__char+ch,PAGE, 1)

main( )

{
int retvals [4] ;

int index ;
SET_M0DE ;

hline(0, 0,219, 80) ; /* draw box around screen */

hline(24, 0,219,80) ; /-* i.e., manual window using nonwindow characters */
verline(0,0,219,24) ;
verline(0,79,219,24) ;
window(3,3,6, 5) ;
window(10, 10,5,6) ;
window(7, 17,10, 11)
window(15,40,10,6)
window(5,50,20,16)

/* read cursor position */
for (index=0; index <4; index++ ) {

curpos(index,index+l) ;

dspio(read_jposition, PAGE, 0,0, retvals) ;

printf("For page PAGE, Row:%u, Col:%u, cursor mode=%o,%o \n",

retvals[3]»8, (retvals[3]«8)»8, retvals[2]»8, (retvals[2]«8)»8)

)

/* get current video state * I
dspio(get_state, PAGE, 0,0, retvals) ;
printf ("Mode: %u, Columns: %u, Active Page: %u \n",

(retvals[0]«8)»8, retvals[0]»8, retvals[l]»8 ) ;

}

window(row,col, width, depth) /* start at row, col, i.e. left upper corner */

char row, col ;

{

/* first do 4 corners *l

curpos(row,col) ;
wrtchar(214) ;
curpos(row+depth,col) ;
wrtchar(211) ;
curpos(row,col+width-l) ;
wrtchar(183) ;

curpos(row+depth,col+width-l) ;
wrtchar(189) ;

/* then connect corners with horizontal and vertical lines */
hline(row,col+l,196,width-2) ;
hline(row+depth,col+l,196,width-2) ;
verline(row+l,col,186,depth-l) ;
verline(row+l,col+width-l , 186,depth-l) ;

}

verline(row, col, ch, len)

char row, col, ch, len ;

(
char cnt ;

for (cnt=len; cnt > 0; cnt — ) {
curpos(row++,col) ;
wrtchar(ch) ;

)
}

Listing 2 (Continued from a preceding page).
56 COMPUTERIANGUAGEB FEBRUARY 1985

functions the return value is stored in the
AX register upon exit from the sub-
program. The other registers are used for
types requiring larger size return values.

For those video I/O Functions that sup-
ply values upon return From interrupt in
AX, no processing is required as they are
passed back to the C caller as the called
functions return value.

Thus, functions read attribute and char-
acter and read dot return information as
the value of the int function dispio upon
return. Get state returns all but the current
active display page as dispio 's int function
value in addition to storing it in the return
area. Get light pen position returns
whether or not the light pen switch is trig-
gered as dispio's char function value or as
its high byte int function value, in addition
to storing it.

In fact, all information can be returned
as the function value of dispio upon return
provided dispio is typed as double. How-
ever, C would have to generate code to
deal with this long float type, which
would result in unnecessary inefficiencies
in most cases.

Efficiency

An instruction efficiency analysis of the
minimum and maximum instruction

sequences for both routines 6 shows that
dispio is almost always faster. When no
values are returned, it executes from 16%
to 30% faster and when values are
returned, from 24% faster to 5% slower.

Dispio is actually faster than the anal-
ysis indicates since it eliminates all excess
pushing of arguments from C. while dspio
requires pushing of filler arguments. For
example, the worst case path for dispio is
when it is doing a read light pen position
and takes 198 cycles. Dispio requires two
arguments from C and thus two pushes,
while dspio would require five arguments
from C or five pushes. Adding the extra
three pushes for dspio yields 2 19 cycles
compared with 198 cycles for dispio ,

Clearly dispio will always be faster than
dspio. If one is willing to put up with
always supplying five arguments to dspio,
the four pushes to restore the stack can be
deleted, in which case it will be almost as
fast as dispio . Of course, since C will be
pushing arguments that aren't needed,
that savings is partly an illusion.

Finally, one could return all values
through C's function mechanism. The
type of the function must then be chosen
carefully, as the type declared will depend
on what values are to be returned. To
some, C opens a Pandora's box since C

requires more knowledge of the machine
and computer-related concepts than most
other languages. To others, C allows
increased productivity resulting from
using a flexible high-level language with
an acceptable efficiency penalty. H

References

1. Backus, John. "The History of FORTRAN
I. II. and III." Tile History of Programming
Languages . Edited by Wexelblat. Richard
L. Academic Press ( 1 98 1 ): 25-45 .

2. Harry, David. "An Implementation
Demonstrating C Portability." Computer
Language (Oct. 1984): 21-29.

3. Microsoft /Lattice C Compiler. Lattice/
Lifeboat (1982). Microsoft (1983).

4 . IBM Personal Computer MACRO Assem-
bler. IBM (1981 1.

5 . IBM Personal Computer Technical Refer-
ence Manual . (Aug. 1981): A-43 through
A-45.

6. Rector, Russell, and Alexy, George. Tile
8086 Book. OSBORNE/McGraw-Hili
(1980).

TITLE DISPIO - SUBROUTINE TO PERFORM DIRECT SCREEN I/O

#define MAX 8

char retvals [MAX] /* used by 3 functions */

/* (1) - read cursor position; retvals[0-l] = current cursor mode */

/* retvals[3] =- col, retvals[2] = row */

/* (2) - read light pen position */

/* retvals[0] = active(l)/inactive(0) */

/* retvals[2-3] = pixel column, retvals[4]=raster line */

/-* retvals[6] = row, retvals[7] = col *•/

/* (3) - get state; retvals[0] = cols, retvals[l] = mode */

/* retvals[2] = page */

setmode - dispio(mode) ;

set cursor type - dispio(set_curs_type, type) ;

set cursor position - dispio(set_cursor,page«8,row<<8+col) ;

read cursor position - dispio(read_cursor_position,page<<8, retvals) ;

read light pen position - dispio(read_j)en, retvals) ;

select active page - dispio(select_j»age+new_page_value) ;

scroll active page up - dispio(scroll_up+number_of_lines, page<<8,

lrow<<8+lcol, rrow<<8+rcol) ;
scroll active page down - dispio(scroll_down+number_of__lines, page<<8,

lrow<<8+lcol,rrow<<8+rcol) ;
get attribute character at current cursor pos - dispio(readac,page<<8) ;

returns attribute character as function value -high byte attribute,

low byte character
write attribute character at current cursor position - dispio(writeac+char ,
page<<8+attribute, repeat count)

Listing 3 (Continued on following page).

; write character - dispio(write_char+character , page<<8, repeat count)

; set color palette - dispio(set_pallette,pallette<<8+color_value) ;

; write dot - dispio(write_dot+color , col, row) ;

; read dot - dispio(read_dot,col,row)

; dot read returned as low byte of function value

; write teletype - dispio(write_teletype+char , (page<<8)+color) ;

; get state - dispio(get_state, retval)

DGROUP GROUP DATA

DATA SEGMENT WORD PUBLIC 'DATA'

ASSUME DS: DGROUP
CASEJTABLE DW SETMODE , CURSTYPE , SETCURS , READCURS , READPEN , SETMODE

DW SCROLL , SCROLL , READAC , WRITEAC , WRITEAC , RE ADAC , WRITEDOT

DW WRITEDOT, READAC, STATE
; case table uses 16 words

RLAYOUT STRUC ; Return address space layout
RAX DW ?
RBX DW ?
RCX DW ?
RDX DW ?
RLAYOUT ENDS
DATA ENDS

PGROUP GROUP PROG

PROG SEGMENT BYTE PUBLIC 'PROG 1

ASSUME CS: PGROUP

PUBLIC DISPIO

Listing 3 (Conri)iued on following page).

QUALITY SOFTWARE AT
REASONABLE PRICES

CP/M Software by

Poor Person Software
Poor Person's Spooler $49.95

All the funclion of a hardware print buffer at a fraction of the
cost. Keyboard control. Spools and prints simultaneously.

Poor Person's Spread Sheet $29.95

Flexible screen formats and BASIC-like language. Prepro-
grammed applications include Real Estate Evaluation.

Poor Person's Spelling Checker $29.95

Simple and fast! 33.000 word dictionary. Checks any CP/M text
file.

aMAZEing Game $29.95

Arcade action for CP/M! Evade goblins and collect treasure.

Crossword Game $39.95

Teach spelling and build vocabulary. Fun and challenging.

Mailing Label Printer $29.95

Select and print labels in many formats.

Window System $29.95

Application control of independent virtual screens.
All products require 56k CP/M 2.2 and are available on B* IBM and 5"
Northstar iormaK, other S" formats add S5 handling charge. California
residents include sales tax.

Poor Person Software

3721 Starr King Circle

Palo Alto. CA 94306

tel 415-493-3735

CP/M ts a registered trademark of Digital ftesearcri

CIRCLE 51 ON READER SERVICE CARD

58 eOMPUTERlANGUAGEBFEBRUARY1°85

The Tools

You Need

T<TC You Thru.

Now the TC/tyvuCT&artt,"

Applications Programmers Toolkit

provides everything you need

to increase your C programming productivity.

AFT" features:

• File Handlers • String Math

• Direct Access • Screen Generator

• Keyed Access • Report Generator

• Generic Terminal Driver • FIFO Que Routines

• String Handling • Source Code

• Manual On Disk (45 pages)

• Tutorial On Disk (33 pages)

• Detailed Brochure on request
UNDC " System 3T Version (available m-ss) . $995

MS-DOS " Version $495

BDS C Version $395

Manual Only* $50

•Manual Coat will be applied, to APT* purchase price within
30 day* (*t'Q restocking charge). U.S. fundn only, ploaao.
TrftdimvLt mmn: UNIX " (AT&T Bell L*b-J. MS-DOS"- (MJCT»eft. lac). BDS C" (BD Sortwm).
Wlunlwu-B - aJjil APT* (Sh*W'. Anurlcvi T*chnotaglM}

Call: (800) 443-0100 Ext. 282

Ask for AFT" or Send Check To:

Shaw w American Technologies

TOijiVtd'WtVU."

^^ 830 South Second Street • Box 648

Louisville, KY 40201, U.S.A.

(C.O.D. and fiirripi Orders AdrJ ¥:> Bhippni^ H m idling J

IMrmm* B*nk Of 1#ui»vtI* CHlu-ni Fidelity B»nk 1.iiili.viI1i- Chunhrr at {VirTiriirn-r

«£OQ(«r. ' ' i TV "■^11T« '" n l^r-?JfcjiSkj»j«£^aj«f ,-

CIRCLE 92 ON READER SERVICE CARD

DISPIO PROC
POP
POP
MOV
XOR
CMP
JA
SHL
JMP

ERROR: MOV

JMP

SETCURS: POP
POP

SETMODE: INT
JMP

READCURS :

SCROLL :
WRITEDOT:

READAC :

POP
INT
POP
MOV
MOV
JMP

POP
POP
POP
INT
JMP

POP
INT
JMP

NEAR

BL,AH

BH.BH

BL.15 ;

ERROR

BX,1

CASE_TABLE[BX]

AX.-l

BX
DX

10H
SI

10H

[DI].RAX,CX

[DI].RBX,DX

10H

INVALID FUNCTION CODE ?

WRITEAC: POP

CURSTYPE: POP

INT

10H

JMP

STATE: INT

10H

POP

MOV

[DI].RAX,AX

MOV

[DI].RBX,BX

JMP

READPEN: INT

10H

POP

MOV

[D1].RAX,AX

MOV

[DIj.RBX.BX

MOV

[DI].RCX,CX

MOV

[DI].RDX,DX

JMP

DISPIO ENDP

PROG ENDS

END

Software
Development

PCDOS/MSDOS

Complete C Compiler

• Full C per K&R

• Inline 8087 or Assembler Floating
Point, Auto Select of 8087

• Full 1Mb Addressing for Code or
Data

• Transcendental Functions

• ROMable Code

■ Register Variables

• Supports Inline Assembler Code

MSDOS 1.1/2.0
Library Support

• All functions from K&R

• All DOS 2.0 Functions

• Auto Select of 1 .1 or 2.0

• Program Chaining Using Exec

• Environment Available to Main

c-window™

Symbolic Debugger

• Source Code Display

• Variable Display & Alteration
Using C Expressions

• Automatic Commands

• Multiple Breakpoints by Function
& Line Number

8088/8086 Assembler

• FAST — Up to 4 times Faster than
IBM Assembler

• Standard Intel Mnemonics

• Compatible with MSDOS Linker

• Supports Full Memory Model

8088 Software Development
Package

199

Includes: C Compiler/Library,
c-window, and Assembler, plus
Source Code fore-systems Print
Utility

c-systems

P.O. Box 3253

Fullerton.CA 92634

714-637-5362

Listing 3 {Continued from preceding page).

CIRCLE 16 ON READER SERVICE CARD

Add EDITING

to your

Software

with

CSE Run-Time

Your program can include all or a portion
of the C Screen Editor (CSE).
CSE includes all of the basics of full
screen editing plus source in C for only
$75. For only $100 more get CSE Run-
Time to cover the first 50 copies that you
distribute.

Use capabilities like Full cursor control,
block move, insert, search/replace or
others. Portability is high for OSes, ter-
minals, and source code.
Call for the "CSE Technical Description"
and for licensing terms and restrictions.
Versions for PCDOS. MSDOS. CPM, more.
Full Refund if /— ► ■ ■ «

not satisfied in -^DOlUtlOfi

first 30 days. „_D UStefTIS

Call 800-821-2492 ns-l Washington Street
CIRCLE 93 ON Nonuell, MA 02061

READER SERVICE CARD 617-659-1571

THE PHOGKAMMEUS SHOP

helps compare evaluate and find products. Get answers.

pSERViCE: FREE LITERATURE

One free call covers all programmer's software. Ask for a
"Packet" on: "AT, BASIC. C. COBOL, Debuggers. Edi-
tors. FORTH. FORTRAN, Libraries. PASCAL. UNIX/PC Or
30 "addons" for "C".

Recent Discovery

Inlrnducing-C : C Interpreter and training system.
Nice. Thorough. PCDOS. Only S95.

'C" Language

MSDDS:CB6- 8087. reliable

Lattice 2.1 - improved - 30 addons

Microsoft C2.x

Williams ■ debugger

instant C interpreter, fast, lull, debug
CPM80: Ecosoft C-now solid, lull, faster
MAC: Megamax - fast, full, light

flunson

PCDOS

8086

lam

BRIEF -Intuitive, flexible
PMATE • powerful

VEDIT- full, liked

ARTIFICIAL INTELLIGENCE

IO LISP- full 100QK RAM PCDOS

TLC LISP -with "classes", nice MSDOS
MicraProlog - by Logic Prog. Assem. MSDOS
PROLOG-86 - standard. Learn fast MSOOS
EXSYS - Expert System PCDOS

OUR
PRICE
call
call
329
call
500
225
295

195
195
119

call
250
285
125
295

FORTRAN

MS Fortran • Improved
DH Fonran-86 - full "77"

F77L - by Leahy -Nice.
RM Fortran

SUPPORT PRODUCTS

Runson

MSDOS

8086

MSDOS

MSOOS

LIBRARIES: BTRIEVE ISAM MSDOS

Clndex+ - ISAM, source, no royalt. 8086

CSHARP Realtime - source, full MSDOS

CUtil by Essential MSDOS

DATABURST - Screens-C. BAS MSDOS

Graphic - 4200 x 3100, source MSDOS

Greenleal C - thorough MSDOS

HALO Graphics ■ last, full PCDOS

TOOLS: MULTIUNK - multitask PCDOS

Polylibrarian-thorough MSDOS

PolyMAKE-manage, compiles PCDOS

Profiler-86-easy to setup, symdois MSDOS

XENIX- "true S3", rich. + C-MSDOS PC

OUR
PRICE

249
259
449
545

215
375
600

139

195
165
175
65

125

1285

PERISCOPE DEBUGGER -
"Reset Box", own RAM.
Registers, symbols,
ine nums. 2 screen,
PCDOS. S295.

Call for a catalog and solid value

800-421-8006

THE PROGRAMMER'S SHOP™

Note: All prices subject to
change without notice.

Mention this ad. Some prices
are specials.

All formats available
Ask about PCs. COD.

CIRCLE 98 ON READER SERVICE CARD

Helper

DIFF and CMP - for "irrtetitgent" file comparisons.
XREF- cross references variables by function and line.
C Flow Chart - shows what functions call each other.
C Beautif ier- make source more regular and readable.
GREP - search for sophisticated patterns in text.
There are several other utilities that help with converting
from one C compiler to another and with printing
programs.

C Helper is written in portable C and includes both full
source code and executable files
for $1 35 lor MS-DOS, CPM-80
or CPM-86. Use VISA,
Master Card or COD.

Call: 800-821-2492

Solution
.Systems

335-1 Washington Street
Norwell. MA 02Q61
617-659-1571
CIRCLE 99 ON READER SERVICE CARD

PROLOG-86

Become Familiar in One Evening

The tutorials combined with the interactive PROLOG-86
Interpreter help you learn the fundamentals of PROLOG
quickly. In a few days you should be able to modify sophisti-
cated sample programs included with the product like:

• an EXPERT SYSTEM

• a NATURAL LANGUAGE INTERFACE.

1 or 2 pages of LOGIC and FACTS create a significant
PROLOG program that might take 1 to 1 5 pages in C.

Prototype quickly in PROLOG. Experiment with applications
that might otherwise require a "definitive requirements
specification".

Programming experience is not required to use PROLOG-
86, but alogical mind is.

PROLOG-86 supports the de facto standard established by
"Programming in PROLOG". Ask aboutthe "Artificial Intelli-
gence Concepts" CONTEST. $1 ,000 prize.
FULL REFUND if not satisfied during first 3 weeks.

Intro Price: $125 for PCDOS, MSDOS or CP/M-86. Most for-
mats available.

For questions/orders, call

800-821-2492
UseVisa,MC,COD.
CIRCLE 97 ON READER SERVICE CARD

-Solution

^Systems

335- l Washington St
Norwell, Mass. 02061
617659.157]

60 COMPUTER LANGUAGE ■ FEBRUARY19S5

EXOTIC LANGUAGE
OF THE MONTH CLUB

magine an inter-
active program-
ming language
where all data is simply a siring of charac-
ters. There are no integers vs. real num-
bers concerns. For instance, the length of
the string "BLUE", the number 3. 14, and
the integer 432 1 are all 4.

Conversion from numeric to string
never takes place: numbers are already
strings! You can use numbers and
alphanumeric strings in any arithmetic
operation. Of course 2 * 3.14 is 6.28. but
"2APPLES" + "30RANGES" = 5.
(Who said you can't mix apples and
oranges!) The rule is that a string used as
a number is converted to the number until
a nonnumeric value is encountered.

Furthermore, even array subscripts are
considered character strings. Thus an
array such as X(J) with a value of 23 is
typical, but how about ACODEfNYC ")
having the value of 212? Of course alpha-
betic and numeric values can co-exist in
(he same array.

Sounds a lot like SNOBOL, but this is
actually a quick start in looking at
MUMPS, a unique language that has been
around since the early 1970s. It has many
rather unusual but very powerful data
management features. To those accus-
tomed to FORTRAN. COBOL, or BASIC
and the like, MUMPS appears quite odd,
plus hard to use and understand. How-
ever, this doesn't have to be the case.

It is really quite easy to work with
MUMPS. In particular, it has enabled the
design of many complex multi-user sys-
tems in which interactive users all share in
the access to and update of data bases.
Before we consider this level of MUMPS,
let's look more closely at some of the lan-
guage's more basic features.

Since all data is represented as a string,
there arc many string related functions, as
in BASIC. For example, to retrieve some
portion of a string you use the EXTRACT
function. Thus

SEXTRACT("Computers",l,J)

returns the /lh through Jlh characters
from the first argument to the function. If

MUMPS: A multi-user, data base language

By J. Edward Volkstorf Jr.

/ is 4 and J is 7 then the function value is
"pute".

Note that functions arc distinguished by
a leading dollar sign (otherwise we would
have designated a three dimensional
array). Variables may be up to eight
alphanumeric characters that begin with a
letter. Other string related functions
include:

■ $LENGTH(X) returns the number of
characters, i.e., the length of variable X

■ $CHARACTER("A ") equal to 65,
returning the ASCII code for a character

■ $FIND("COMPUTE", "MP") has the
value 3, finding the first occurrence of the
string "MP" in the first argument
COMPUTE

■ $PIECEC1ST,2ND,3RD"","N) returns
the Nth comma delimited portion of the
first string argument, such that if N is 2
then 2ND is returned. The second argu-
ment, the comma here, determines the
delimiter

Of course functions can be nested, and
numeric values can be used as arguments.
For example,

■ $FIND(N, ". "J will be either if N is an
integer (i.e., no decimal point) or some
value greater than if it is a real number

■ $LENCTH($PIECE(NAME, ","1))
might be used to calculate the length of
someone's first name, those characters
preceding the first comma in the string
NAME

As mentioned initially, arrays may also
contain strings. They may be any dimen-
sion, and subscripts may be ASCII char-
acter strings. A typical array might be
used, as in most languages, to store a table
of numbers:

X(1) = 12,X{2) = 23,X(3) = 34,etc.

but since MUMPS allows those subscripts
to be character strings you can design data
tables that look like:

ACODE("SAN DIEGO") = 619
ACODE("LOSANGELES") = 213
ACODE("SACRAMENTO") = 916

where the array ACODE defines the valid
telephone area code numbers for a set of
cities. For data such as a person's name,
street address, arbitrary words, and the

like, this allows some incredible savings
in program code when manipulating data.
But how do you retrieve the values?

In the first example the array A' could
use a subscript variable that ranges from I
to 3. MUMPS handles the area code array
differently, using the SNEXT, BORDER,
and SDEFINE functions.

Both $NEXT and SORDER take a sub-
scripted variable and determine the sub-
script value that follows in that array
based on the MUMPS collating system.
(Note that this gets tricky, since for
numeric subscripts 123 follows 45 which
follows 6, but for a strict string collating
sequence, left to right character com-
parisons are made: the sequence would be
123, 45, then 6). MUMPS determines
whether subscripts are collated as numer-
ics or as ASCII character strings .

Thus for ACODE, the following holds:

■ $NEXT(A CODE( ' X OS ANGELES ' 'J ;
would be "SAN DIEGO"

■ SNEXT(ACODE("SA ")) would be
"SACRAMENTO"

■ SNEXT(ACODE("SAN DIEGO")) is

- 1 , indicating that no further subscripted
values were found.

The last example is where $NEXT and
SORDER differ. The latter returns a null
string instead of the — 1 .

There is another way to determine if a
single value exists using the SDATA func-
tion. It takes any variable name as an
argument and returns 1 if the variable is
defined , a 10 if it is a pointer to a lower
level (such as the first subscript in a two
dimensional array), or 1 1 . which covers
both cases. Otherwise a zero is returned.
Thus,

■ SDEFINE(ACODE("S4N DIEGO"))
returns 1

■ $DEFINE(ACODE("SANTA CLARA"))
returns

■ $DEFINE(ACODE) returns 10

■ but $DEF1NE(AREAC("L0S ANGE-
LES ")) returns since AREAC is not the
correct array name (assuming it is not
used).

ong with this set
of string func-
tions is a host of

numeric and string operators. Many are

very powerful in nature. For numerics,
the typical arithmetic operators of + , — ,
*, / are used for addition, subtraction,
multiplication and division. Also an
integer divide is symbolized by \ and
modulo operator by #. The last two oper-
ators make for very concise expressions
such as

123 \ 10= 12
123#10 = 3
123 \ 10 # 10 = 2

For string operations, the underline
designates concatenation. The right
bracket is the contains operator which
tests whether the first argument contains
the second. The left bracket signifies a
string follows comparison. As you might
expect, the usual Boolean operators, =,
< , > , < = , and > = , are used for
assorted equality and inequality checks.

Keeping track of operator precedence is
simple in MUMPS; there isn't any. Oper-
ators arc processed in a strict left to right
sequence except for expressions in paren-
theses which, as expected, are processed
first.

One of the more powerful Boolean
operators is the ?, which tests for pattern

GOOD NEWS!

HIT W. Virginia si.
Milwaukee. H7 .ViL'OI
(414) 276-2937

for the

6809

MAS NEVER

BETTER!

INTROL-C/6809,Version1.5

Introl's highly acclaimed 6809 C
compilers and cross-compilers are now
more powerful than ever!

We've incorporated a totally new 6809
Relocating Assembler. Linker and Loader.
Initializer support has been added, leaving
only bitfield-type structure members and
doubles lacking from a 100% full K&R
implementation. The Runtime Library has
been expanded and the Library Manager is
even more versatile and convenient to use.
Best of all, compiled code is just as
compact and fast-executing as ever - and
even a bit more so! A compatible macro
assembler, as well as source for the full
Runtime Library, are available as extra-cost
options.

Resident compilers are available under
Uniilex, Flex and OSS.
Cross-compilers are available for PDP-
11/UNIX and IBM PC/PC DOS hosts.

Trademarks:

!ntrol-C. Introl Corporation

Flex and Uniflex. Technical Systems Consultants

OS9, Microware Systems

PDP-11. Digital Equipment Corp.

UNIX, Bell Laboratories

IBM PC. International Business Machines

For further information, please call or write.

matching. A 1 or truth value is returned
depending on whether or not the first
argument follows the pattern designated
by the second argument. Patterns are
determined by various character class
codes and string constants, which can be
modified by a numeric repeat count. For
example, in the following statement the
letter N designates numeric digits that test
whether the variable SSN is a valid social
security number:

IFSSN?3N1"-"2N1"-"4N

Thus the pattern is three digits, a dash,
two more digits, a dash and ending with
four digits.

In the following example, the letter V
indicates upper case alphabetic characters
and A , an alphabetic character. The
period is used to mean any number of
characters as a repeat count. The
expression

IFNAME?1U.A1", "1U.A

might be used to check for a person's
name, formatted as a last name and first
name separated by a comma .

Other letter codes include C for control
characters. E for any character (every-
thing), and P for punctuation.

Logical expressions are combined with
& for and and .' for or. The apostrophe
may be used in front of any logical oper-
ator for the inverse of the operator. Thus
IFX<0 !(X>1) tests whether X is less
than zero or if it is greater than one, and
IFM'I"." & (M?.N) tests whether M is
all digits and does not have a decimal
point.

Another powerful operator is @. which
represents indirect reference of "indi-
rection." Indirection is a unary operation
that is performed on character strings.
The result of an indirection operation
depends on how it is used. Basically, @
takes a value and treats it like a variable:

SET NAM = "X" ; variable NAM gets
value "X"

SET ©NAM = 2 ; indirectly sets vari-
able X to value 2

or an expression:

SET ARG = "X = 3,Y = 4"; define an

argument
SET @ARG ; sets X to 3 and Y to 4

Indirection is useful in processing differ-
ently named arrays with one statement.

sin BASIC,

input and output
in MUMPS is
oriented around interactive video display
and printer terminals. The command

CIRCLE 32 ON READER SERVICE CARD
62 COMPUTER LANGUAGES FEBRUARY 19B5

WRITE outputs the values for constants,
variables, and expressions to the currently
assigned I/O device.

Input is done in a similar manner using
the READ command. For example, the
code in Listing 1 reads a set of integer val-
ues and prints their sum. Note the pattern
matching command and the entry of a null
string "" to indicate no more data. Com-
ments arc denoted in MUMPS as all text
that follows a semicolon on a single line.

MUMPS program lines begin with one
or more spaces or with an alphanumeric
line label. Statements are separated by a
single blank space. Since the syntax of
each statement determines its format and
overall length, statement delimiters are
not needed. Control may be passed to
named lines with the GOTO or DO
commands.

The S£7"command assigns values to
any number of variables. Note that the IF
command does not use a THEN. MUMPS
continues processing with the command
that follows the logical expression if it is
true. Otherwise control is transferred to
the beginning of the next line.

An ELSE command is available but not
in the sense of most other languages .
Instead MUMPS has an internal last truth
value that may be checked at anytime with
the ELSE.

Thus an IF might have a corresponding
ELSE located many lines of code away.

The IF can test this internal truth value by
using a null expression, designated by two
blank spaces after the IF. This means the
results of an IF evaluation can be checked
in other areas of a program without the
original expression being coded.

Loops in MUMPS are coded using the
FOR command. FOR repeatedly sets a
variable to some value based on one of the
four following basic formats:

FORI = l:l:10;setsltol,2,3... 10

FORN = "TOM","DICK","HARRY";

sets N to each of the strings that

follow
FOR 1 = 1:1 UNTIL I >99; increments I

until the condition following the

UNTIL is true

FOR 1 = 1:1 WHILE I < 100;

increments as long as the condi-
tion after the WHILE is true

The scope of the FOR is only to the end
of the line. Thus the following counts how
many numbers are greater than in the
100 element array NUM:

SETGZ = 0FORI = lTO100IF

NUM(I)>0SETGZ = GZ+1

The DO command calls subroutines. It
is like the GOTO in all respects except
that anywhere in the code to which control
is transferred, a QUIT statement can be

SET TOTAL=0,CNT=0 ; initialize total and count variables

IPN READ "Enter a number: ",NUMB

IF NUMB="" GOTO DONE ; exit on a null entry
IF NUMB'?.N WRITE NUMB," is not valid" ; validate number
ELSE SET TOTAL=TOTAL+NUMB,CNT=CNT+l ; update values
GOTO IPN ; get next number

DONE WRITE "Read ",CNT, "numbers for a total of ".TOTAL
HALT ; end of program

Listing 1.

OASYS

TOOLKIT
SNAP-SHOT

WIZARD

C-8086

CROSS AND NATIVE OPTIMIZING COMPILERS
FOR 8086/87/88/186/286

As part of OASYS' "ONE STOP SHOPPING" service for software engineering tools, we are proud to announce the addition
of WIZARD C to our integrated collection of more than 50 professional programming tools (e.g. compilers, assemblers,
linkers, debuggers, simulators & translators) for M68000, Intel 8086/80186 and NS32000 micros.

WIZARD C benchmarks (against Lattice, Microsoft) prove that it is, by far, the most advanced, full featured, fastest,
tightest, optimizing C compiler now available for cross and native (PC) development Here's why . . .

FEATURES

• Complete K&R implementation plus
V7, III extensions

• Supports 8087 Floating Point

• Built-in LINT

• Long, medium, short models

• 190+ UNIX 111 functions — complete
run-time library

• In-line assembly allowed

■ 100+ extensive warnings/diagnostics

• Intel and Microsoft compatibility at
source and object levels

■ Written in C — easily ported

• Comprehensive bound documentation

• Supports DOS 2.0, 2.1 , IBM/BIOS

W86CH-11/84

SUPPORT TOOLS

• Symbolic C Source

Level Debugger (CDEBUG™)

■ 1 00% Intel compatible structured
Macro Cross Assembler,
Linker/Locate rand Librarian

• 8086 Simulator

• Floating point math package
(40+ functions)

• C Time Profiler (CLUE™)

• Checkout compiler (SAFE-C")

• Comm. utilities for up/down
loading to MDS, TEK, Microtek

AVAILABILITY

NATIVE: PC/XT/AT using

MS/DOS, PC/DOS (Xenix soon)

CROSS: VAX/VMS, Bsd 4. 1 , 4.2, III, V;
8086's, 68000's (All Unisoft III, V;
ports); Cailan, Masscomp, Sun,
Pyramid, dozens more . . .

Call for pricing, OEM, Site, Corporate,
Source and Maintenance licensing
information.

OASYS

60 ABERDEEN AVENUE

CAMBRIDGE, MA 021 38

(617)491-4180

CIRCLE 75 ON READER SERVICE CARD

executed causing a return to the next
expression in the calling DO statement. A
DO may call one or more subroutines
such as:

DO SUB1 ,SUB2,SUB3 ; perform in
succession the 3 subroutines

Related in some respects to indirection
is the XECUTE command . It evaluates its
argument as a command string to be exe-
cuted as though it were explicitly coded in
the program. For example.

XECUTE $PIECE("GOTO NEXT/DO
PREV",'7",N)

either does a GOTO NEXT or a DO PREV
depending on whether N has the value of )
or 2. Obviously some very cryptic and
run-time dependent code can be created
with this command.

A noteworthy coding format in
MUMPS allows conditional execution of
all commands, with minor exceptions. A
command name may be followed by a
colon and some expression. The expres-
sion is evaluated and. if true, the com-
mand is processed. Otherwise it is
bypassed. Thus the FOR loop counting

example could be coded

SETGZ = 0FORI = 1TO100
SET:NUM(I)>0GZ = GZ+1

A GOTO or DO command also may use
this conditional execution as in:

DO:X » "YES" SAVE ; call subroutine

SAVE if X is YES
GOTO:!' = SUM ; goto line labeled

SUMiflisnotO

The expression can also follow argu-
ments to the DO and GOTO, which is
handy for selectively executing certain
subroutines:

;

'• .«

How Do You
Measure A Good
C Compiler?

CODE SPEED & SIZE

The Laltice C ComDiler
"generates code thai is
quite compact and last
running!' Peter Norton,
PC Magazine

CONSISTENT RELIABILITY

"The Laltice Compiler has
performed reliably and
predictably" ft Phraner,
Byle Magazine

COMPILE TIME

"Lattice is a real performer!'
Houston. Brodrick,
Ken!, Byte Magazine

THIRD-PARTY LIBRARIES
More than 40 library products
are currently available
for Lattice

UNIX V COMPATIBILITY

The Lattice Library is
UNIX Vcompalible

DEBUGGER SUPPORT

The Lattice C-SPRITE
Debugger is now available

DOCUMENTATION

Laltice "is thorough
and excellent! D. Clapp,
PC Magazine

UNIX is a trademark of
AT&T Bell Laboratories

UPDATE POLICIES

Lattice provides Iree bugfix
updates for 90 days

COOPERATING PRODUCTS

New LMK Utility, dBC Library,
CVUE Screen Text Editor, CURSES
Screen Library and GSS Graphics
are available Irom Lattice

VENDOR REPUTATION

Lattice is used in more
commercial products than any
other C Compiler, No run-
time license is required.

ALL MEMORY MODELS

Lattice C has 7 memory models
available to allow the best
solution for the task at hand

AVAILABILITY OF
CROSS COMPILERS

SATISFACTION GUARANTEED!

Ask About Our "Trade Up To Lattice C" Policy

01 LATTICE

Lattice', Inc.
P.O. Box 3072
GlenEllyn, IL 60138
(312)858-7950
TWX 910-291-2190

SUB1:X<0,SUB2:X = O,SUB3:X
>0,SUB4

which performs one of the subroutines
SUB1 . SUB2 . or SUB3 depending on the
value of X and always calls SUB4.

y now many
ireaders may
'think MUMPS is
a variation of BASIC but without line
numbers and numeric and string data typ-
ing. To some extent this is true. Where
MUMPS truly leaves all other languages
behind is in its ability to process complex
data structures that are disk resident.

MUMPS files do not exist in the tradi-
tional sense. All disk data is a string of
ASCII characters, named in the same
manner as local memory-resident vari-
ables, with the exception of a A (carat)
preceding the name. The significance of
this is that all statements, commands,
functions, operators, etc may use disk
variables in a manner identical to local
variables.

MUMPS calls its disk data globals
since disk data may be shared by all users.
Memory-resident variables, on the other
hand, are local to one particular MUMPS
process. Thus instead of opening, read-
ing, and writing a file, you simply S£7'the
global to some value. For instance, the
following statement adds one to a status
global and then displays its current value:

SETASTAT = ASTAT+1 WRITE
is",ASTAT

Status

CIRCLE 53 ON READER SERVICE CARD

64 COMPUTER LANGUAGE" FEBRUARY 19B5

The SDEEINE function, mentioned ear-
lier, determines whether a global exists or
not. Thus a statement like the one below
might precede the above update of global
STAT

IF $DEFINE(*STAT) = WRITE "Status
not defined"

These two examples illustrate the sim-
plest forms for using a global. More often
globais arc processed as multi-
dimensional arrays where the subscripts

are either numeric or string subscripts.
Processing a global then takes on the form
of array processing using the FOR loop
and SNEXT or SORDER functions.

A strict numeric sequence of integers
might be used as subscripts to store a
series of statistical values, where each
value in the global is a single number. For
example, let the global A X contain 200
values. Then their sum is simply:

SETTOTAL = FOR 1 = 1:1:200
TOTAL == TOTAL + AX(I)

which is identical to how the values would
be totalled if ^X were a local array.

As with local arrays, global subscripts
are also defined sparsely, i.e., only those
values defined are actually stored in the
global. An accounting system could use a
global to track the status of purchase
orders where the subscript is a purchase
order number. The data stored at a partic-
ular node in the global might contain
items such as the date ordered, the cus-
tomer number, and so on. For further
detail on the data— the purchase order in
this example— a second set of subscripts
might be used for information about the
nature of the order, the line items in it,
terms, etc.

We could have a situation where the
code presented in Listing 2 would tell a
user if a purchase order exists and the date
when it was made based on the first sub-
script in global ^PONUMB.

A second level of subscripts might be
the line item numbers in the PO. with a
subscript of zero indicating how many
lines, plus other related information.
When a PO is confirmed to exist using the
previous example, the line items might be
printed, unformatted, as in Listing 3.

The unformatted data is noted since the
values in the string would usually be vari-

able length with some special character
delimiter. The SPIECE function is then
used to determine individual field values.
A line item node in the above global might
look like:

1 2 ^ MX! 02 A DS/DD Diskettes A 1 .50
A 18.00

which could indicate 12 model MX 102
DS/DD diskettes at SI .50 each for a total
of S18.00. The A is an often-used
MUMPS field delimiter.

PON READ "Enter PO number: ",P0

IF P0*?.N WRITE "Invalid number" GOTO PON
IF $DEFINE( A PONUMB(P0))=0 WRITE "Not on file"
ELSE WRITE "Ordered: "/PONUMB(PO)
GOTO PON

Listing 2.

SET LN="P0NUMB(P0,0)

WRITE LN," line items for P0 ",P0

FOR I=1:1:LN WRITE !,"Line ",1," ",

Listing 3.

'P0NUMB(P0,I)

OASYS

TOOLKIT
SNAP-SHOT

C-68000

CROSS AND NATIVE OPTIMIZING COMPILERS

FOR 68000/10 (and 68020 SOON)

OASYS offers a "ONE STOP SHOPPING" service for software developers in need of proven B-, 16- and 32-bit cross and
native tools for Unix and non-Unix 68000, 8086 and 32000 systems. Our critically acclaimed and widely used 68000 tool kit
offers high quality, reliable, cost-effective tools.

The OASYS 68000 tool kit consists of Green Hills compilers (C, Pascal and FORTRAN), our own M68000 Macro Assembly
Development package, and dozens of other OASYS compatible support tools. Simply stated, we beat the competition on
price, speed and tightness of emitted code.

C-68000/10

• Full K & R with Western Electric and Berkeley
extensions

• Complete run-time library available as
source. No royalty if passed on.

• Supports DEC & IEEE Floating Point

• Integrated optimizer: 30% tighter code than
Portable C; 4 times faster

• Generates M.I.T. or EXORmacs assembly
source code

• Interfaces to all OASYS tools and Pascal,
FORTRAN and PL/M-68K compilers

• Ideal for cross development of boards with no
OS, a kernel OS (e.g. VRTX, PSOS, MTOS),
or Unix based 68000's

C68K-1 1 1SA

68000/1 Assembly package Other tools

• EXORmacs compatibile Macro
Assembler, Linker, Librarian, and
Cross Reference Utility

• Generates S-records and a.o.

• PIC and reentrant code

• Used 2 years in house

• Over 3,000 sold to date

• Runs on VAX, Prime, PDP-11,
68000's, 8086/88 (PC)

• Written entirely in C
Coming soon

• 68020 C and Cross assembler

Symbolic C Source Debugger

68000 Simulator
& Disassembler

C Linecount and Time Profiler
Utility (CLUE TU )

LINT for VAX/VMS

Check Out compiler
(SAFE-C ,U )

Communications tools

OASYS

60 ABERDEEN AVENUE

CAMBRIDGE, MA 02138

(617)491-4180

CIRCLE 80 ON READER SERVICE CARD

NGS FORTH

A FAST FORTH
OPTIMIZED FOR THE IBM
PERSONAL COMPUTER
AND MSDOS COMPATIBLES.

♦79 STANDARD

*FIG LOOKALIKE MODE

♦PC -DOS COMPATIBLE

♦ON-LINE CONFIGURABLE

♦ENVIRONMENT SAVE
&LOAD

♦MULTI-SEGMENTED

♦EXTENDED ADDRESSING

♦AUTO LOAD SCREEN BOOT

♦LINE AND SCREEN EDITORS

♦DECOMPILER &
DEBUGGING AIDS

♦8088 ASSEMBLER

♦BASIC GRAPHICS &, SOUND

♦NGS ENHANCEMENTS

♦DETAILED MANUAL

♦INEXPENSIVE UPGRADES

♦NGS USER NEWSLETTER

A COMPLETE FORTH
DEVELOPMENT SYSTEM.

PRICE: $70

PLEASE INCLUDE 12 POSTAGE k
HANDLING WITH EACH ORDER.

CALIFORNIA RESIDENTS :
INCLUDE 6.5% SALES TAX.

NEXT GENERATION SYSTEMS

P.O.BOX 2987

SANTA CLARA, CA. 9 5055

(408) 241-5909

MUMPS provides a syntactic con-
venience in processing giobals. called the
naked reference, II essentially means that
a global reference may be made without a
name and the last level in the global will
continue to be processed. This means the
global coded *PONUMB(PO,I) could
have been coded as simply A (I) since the
initial reference was in the first line at the
second level subscripts.

With two or more subscripts used to
define their data, giobals take on the form
of tree structures. Although not a data
base management system in the fullest
sense. MUMPS provides all the language
features needed for creating any hier-
archical data base.

Most MUMPS systems implement their
global data structures using some form of
balanced B* or multiway tree structure.
This guarantees a reasonable response
time to examine any node in any size
global.

UMPSisalsoa
multi-user data
management
system. Its global data structures may be
referenced by any user on the system. For
instance, the purchase order global could
be referenced by any number of users on a
system checking the status of their pur-
chases. The typical MUMPS system has
all global references occurring in a buffer
pool (sometimes called a disk cache) so
that each user has access to the same and
most current data.

Updating a file is another matter. To
control the sequence in which a global is
updated there is a LOCK command. This
allows a user to have exclusive control of
an entry in a lock table that in turn typi-
cally defines an area of a global.

In the purchase order example, a
receiving area might be changing the sta-
tus of some line items in the file. In the
clerical area a buyer might also change
something in the PO that is on back-order.
Either user could have their program
LOCK the entire order if both used a state-
ment of the form:

LOCKApoNUMB(PO)

On the other hand, if the changes are

only specific to one line item in the order
then the control may be:

LOCKApoNUM8(PO,LINE)

where PO is the purchase number and
LINE the line number being changed .

Users on the system making inquiries
do not have to use the locking code since
they will always be retrieving the most
current PO data from the buffer pool.

In general, access codes, passwords,
and the like on commercial MUMPS sys-
tems provide an overall control mech-
anism. They usually map a group of
related giobals to a certain user class

when those users sign on to the system.
Many variations exist, even to the extent
of coding certain giobals as read only, or
read/write by user class.

The HANG and JOB commands are also
useful in a multi-user environment. The
HANG command suspends program exe-
cution for a specified number of seconds.
This might be used in time dependent pro-
cesses, such as waiting until a certain time
of day has arrived before continuing with
program execution.

The JOB command initiates another
MUMPS process. This might be some-
thing as simple as a print spooler to a pro-
cess that files the records generated by a
number of other data entry processes.
With the HANG command the second pro-
cess could be set to function every few
minutes or so instead of continually
wailing for records to file.

t this point I
would like to
describe the

typical MUMPS system layout. A single-
user system consists of basically the inter-
preter and the single-user area. The user
area has a fixed space that describes the
current context in which the user is func-
tioning. Three variably sized areas also
exist: the user's program code, the local
variables (the symbol table), and a stack
area for intermediate values during pro-
gram execution. (Figure 1.)

For multi-user systems, an executive
handles the time sharing and context
switching between users. In the disk
buffer pool all global references arc han-
dled. (Figure 2.)

This area is also used to hold program
routines called in from the disk by the DO
and GOTO commands. Some venders
even share program code between users.
Thus the user partition only consists of the
fixed context area, the symbol table, and
stack.

Some systems do not fluctuate just in a
fixed partition area and thus allow the
symbol area to be completely dynamic in
size. Some systems even swap partition
overflows to and from the disk. In these,
MUMPS becomes a virtual machine much
in the same manner as the IBM 360/370
OS.

Some MUMPS vendors who offer net-
worked systems provide another very
sophisticated area of global processing. In
these systems a global has some additional
coding within its name designating that it
resides on another machine. For example,
some vendors use brackets to indicate the
networked machine as in

SET

A [MACH3]GLOBAL(SUBl,SUB2)
=

CIRCLE 74 ON READER SERVICE CARD

66 COMPUTERLANGUAGEBFEBRUARY19B5

UniPress

Product

UPDATE

AMSTERDAM COMPILER KIT
FOR UNIX

Package of C and Pascal compilers, cross
compilers and assemblers. The Kit was
developed at the Vrije University of Amsterdam
and has been used in Europe for years.
Available for a wide variety of host and target
machines: Unix 68000, VAX and PDP-1 1 and can
produce code for any of those machines, plus
the 8086.

Collection of programs designed to simplify the
task of producing portable compilers and
interpreters.

The Kit contains complete internals
documentation describing how to make

modifications needed to add a new program

language or a new target machine.

Cross assemblers are also provided for 8080,

Z80, Z8000, 8086, 6800, 6809, 68000, 6502, and

PDP-1 1. Cross interpreters, for testing, are also

included.

The Kit consists of 8 components: preprocessor,

front end, peephole optimizer, global optimizer,

back end, target machine optimizer, universal

assembler, utility package.

The Kit includes all sources and documentation

for VAX/Unix 4.1/4.2, PDP- 1 1 /V7, and MC68000

implementations.

OEM terms available • Much more Unix software,
too! • CaJl or write for more information.

Source

Price: Full system $9950

Educational Institution $995

Source requires an AT&T source compiler UfliPfGSS Sof tlUOfO, IflC.

license
Binary

Full system $4995
(contact us for specific
availability)

2025 Lincoln Highway, Edison, N] 08817
201-985-8000 • Order Desk: 800-222-0550 (outside NT)
Telex 709418 • Mastercard and Visa

Lattice is a registered trademark of Lattice. Inc. Unix is a trademark
of Bell Laboratories. MS-DOS is a trademark ot Microsoft.

CIRCLE 38 ON READER SERVICE CARD

Programming
Guidelines

Thomas Ptum

C LANGUAGE PROGRAMMING

From Plum Hall. ..the experts in C training

Learning to Program in C 372 pp., th- x w. Price S25 00
A practical step-by-step guide for everyone acquainted with com-
puters who wants to master this powerful "implementer's language"
Inside, you will learn how to write portable programs for the full
spectrum of processors, micro, mini and mainframe

C Programming Guidelines wo pp.. iw x 10-, Price S25.00
A compilation of standards for consistant style and usage of C
language. Arranged in manual page format for easy reference, it
presents time-tested rules for program readability and portability.

FREE

C LANGUAGE POCKET GUIDE!

A handy C language programming
pocket guide is yours free when you order
either (or both) of the manuals above
A full 14 pages of valuable C language
information!

PLUM HALL

1 Spruce Av, Cardiff NJ 08232
Please send me:

The experts in C and UNIX™ training.
Phone orders: 609-927-3770

information on C and UNIX Training Seminars
copies of Learning to Program in C @ S25.00taopy
copies of C Programming Guidelines @ S25.00/copy

NJ residents add 6% sales to<

COMPANY.
ADDRESS _

CITYfSWEIZtP

i Check American Express

CARD *

UNIX ii a irooemcfi oi ATM Ben Laoo'oione>

Master Cord

EXP. DATE

Signatute,

CIRCLE 50 ON READER SERVICE CARD

WHY WOULD ANY SANE PERSON

SPEND $199 FOR A BetterBASIC

SYSTEM WHEN DOS's IS FREE?

HERE ARE 10 REASONS: TEST YOUR SANITY

I. Full support for 640K memory 2* Structured language with BASIC syntax
3* Separately compiled program modules *L Speed: FAST 5* Extensibility (Make
your own BASIC.) 0« User-defined procedures and functions /• Built-in windows
support O. Interactive programming language based on an incremental compiler
9. 8087 math support 10. Runs on IBM PC, IBM PC/XT and compatibles

Summit Software

Technology, Inc.™

P.O. Box 99, Babson Park

Wellesley, MA 02157

(617) 235-0729

NOW AVAILABLE FOR
THE TANDY 2000 & 1200

BASIC

Sane Programmers Order BetterBASIC Now

Price: $199 8087 Math Module: $99
Runtime System: $250 Sample Disk: $10

BeilerBASIC is o trademark □( Suaimi! Serf two re Technology, (nc

IBM PC. IBM PC/XT end PC/DOS ate trademarks of Iniernofional
Business Machines Corp. MS-DOS is a trademark of Microsoft! Corp.

MasterCard, VISA, RO. Checks, Money Orders,
ondCO.D. accepted.

CIRCLE 88 ON READER SERVICE CARD

$1 30 while quantity lasts.

work < | ' rvvdrffl V^^

ie Intel ceramic ^rW '■'■ * F

/hich operates at +J^BJr\

Put the
speed and power
of the 8087 to
for you.

We've got the
8087-3 chip which
5 MHz.

Order by phone and we ship the same day
for Visa or MasterCard customers. Or you can send cash,
check or M.O. (Sorry, no COD's). Add California state
tax, if applicable and S3 for shipping in the U.SA or
Canada; $15 for foreign air mail. (415) 827-4321

_. _ . . VISA MASTERCARD

Steve Rank, Inc.

1260 Monument Blvd., Concord, CA 94518

CIRCLE 87 ON READER SERVICE CARD
68 COMPUTER LANGUAGE "FEBRUARY 1985

CIRCLE 90 ON READER SERVICE CARD

A MUMPS program that SETs the
global value is actually passing data
through a high-speed communications
device to another MUMPS computer's
disk. Once again, other than the slight
name change, the remote global is pro-
cessed in the same manner as one on the
current machine, which is the same as
locally defined variable data.

As you can sec now. MUMPS is not just
a language but a unique way of developing
multi-user data base oriented systems on
mini, and microcomputers. When 1 think
of MUMPS I think of the fanaticism of C
or Forth. Like these languages. MUMPS
is fostered by a dedicated set of program-
mers who often work with systems con-
taining over 100 interactive terminals that
are updating and accessing the same data
base on moderately priced
minicomputers.

MUMPS has its roots in the health care
environment; in 1969, the initial version
of the language was created at Massachu-
setts General Hospital. The following
seven years saw a small proliferation of
various dialects which culminated in an
ANSI standardized version that was first
offered by a number of hardware manu-
facturers, most notably Digital Equip-
ment Corp. on its PDP-1 1 .

The language is currently available on a
wide range of minis, including the Data
General Eclipses, Prime, Tandem, and

others, MUMPS is also available in a
multi-user environment on the IBM PC
and 6809-based microprocessors. Single-
user systems are also available for the
Apple, 6809. IBM PC, and 8080 systems.

MUMPS has an enormous number of
applications in the health care field, par-
ticularly in laboratory and radiology data
management where there is extensive
variable length textual data.

Another important application includes
overall patient data management for
admissions, business office and medical
records. Many clinic and physician group
practice systems also use MUMPS and of
course many business-oriented applica-
tions are written in it.

The key element MUMPS offers is its
ease in developing multi-user data man-
agement systems. In the health care appli-
cation examples noted, it is essential to
allow all users access to individual patient
data. For example, in a laboratory appli-
cation, test results can be viewed as soon
as they are entered into the system on
video display terminals at the nursing sta-
tion, physicians' lounge, or even the phy-
sician's office if he or she has a modem
for dialing into the hospital's computer.

For further information contact the
MUMPS Users Group. They offer a wide
assortment of publications and services
for the novice and veteran MUMPS user.
A national meeting takes place every year
in June in different parts of the country.
For more specific information on actual

commercial MUMPS systems, the MUG
provides a list of current vendors of
MUMPS language processors and
MUMPS applications. Four
microprocessor-based MUMPS system
vendors arc also included in the
references, JJ

References

1. MUMPS User's Group, 4321 Hartwick
Rd„ Suite 308. College Park. Md. 20740

2. Micronetics Design Corp.. 932 Hungerford
Dr., #\ 1, Rockville, Md. 20850, (301)
424-4870. Motorola 6809. 68000. IBM PC
MUMPS implementations.

3. Richard Walters. Ph.D.. Dept. of Com-
puter Science. Univ. of California. Davis.
Calif. 95616. Public domain 8080. 8088.
Z80 MUMPS.

4. Innovative Computer Systems Inc., 1660 S.
Albion, Ste. 902. Denver, Colo. 80222.
(3031 759-7633. IBM PC implementation.

5. Eclectic Systems Corp., 16260Midway
Rd . , Addison . Te xas 7500 1 . (2 1 4)
661-1370. Various microprocessor imple-
mentations of MUMPS.

J. Edward VolkstorfJr. is a free-lance soft-
ware consultant living in Chesapeake, Va.

Single user MUMPS system memory layout Multi-user MUMPS system memory organization

Interpreter

Fixed area
Program area
Stack
Symbol table

Time sharing executive
Interpretor

Disk cache/buffer pool

User

partition

areas

Figure 1.

Figure 2.

wizard C

Fast compiles, fast code and great diagnostics
make Wizard C unbeatable on MSDOS. Discover
the powers of Wizard C:

ALL UNIX SYSTEM 111 LANGUAGE FEATURES.

UP TO A MEGABYTE OF CODE OR DATA.

SUPPORT FOR 8087 AND 80186.

FULL LIBRARY SOURCE CODE, OVER 200 FUNCTIONS.

CROSS-FILE CHECKS OF PARAMETER PASSING.

USES MSDOS LiNK OR PLINK-86.

CAN CALL OR BE CALLED BY PASCAL ROUTINES.

IN-LINE ASSEMBLY LANGUAGE.

240 PAGE MANUAL WITH INDEX.

NO LICENSE FEE FOR COMPILED PROGRAMS.

The new standard for C Compilers on MSDOS!

Only $450

WSS

For more informatior. call (617) 641-2379

Wizard Systems Software, Inc.

11 Willow Ct., Arlington. MA 02174

Visa/Mastercard accepted

CIRCLE 86 ON READER SERVICE CARD

•

MACWTOSHISA
P£GSTEHH>TPADEMAPJ<
OF APPLE COWLTTEflNC

RP/

T.M.

By the author of Hayden's "CP/M Revealed."

New resident console processor RCP and new
resident disk operating system RDOS replace CCP
and BDOS without TPA size change.

User files common to all users; user number
visible in system prompt; file first extent size
and user assignment displayed by DIR; cross-drive
command file search; paged TYPE display with
selectable page size. SUBMIT runs on any drive
with multiple command files conditionally invoked
by CALL. Automatic disk flaw processing isolates
unuseable sectors. For high capacity disk systems
RDOS can provide instantaneous directory access
and delete redundant nondismountable disk logins.
RPMPIP utility copies files, optionally prompts
for confirmation during copy-all, compares files,
archives large files to multiple floppy disks.
RPMCEN and GETRPM self-install RP/M on any
computer currently running CP/M®2.2. Source
program assembly listings of RCP and RDOS appear
in the RP/M user's manual,

RP/M manual with RPMGEN.COM and GETRPM.COM
plus our RPMP1P.COM and other RP/M utilities on
8"SSSD$75. Shipping $5 ($10nonUS). MC,VISA.

118 SW First St. - Box G
Warrenton, OR. 971U6

icro •

ethods, Inc.

(503) 861-1 765

CIRCLE 52 ON READER SERVICE CARD

CIRCLE 13 ON READER SERVICE CARD
70 COMPUTER LANGUAGE ■ FEBRUARY 1985

YOU DON'T NEED

EUREKA! ??

CONGRA TULA TIONSU

Wc admire your talents. After all, few people can
remember where tn find that six month old letter to Wonder
Waffle Works, or which of the twenty versions of
IMPORTNT.BAS istheoneyou need yesterday.

Or maybe we should envy your spare time. Ah, to be able lo
haul out a slack of disks, slip each one into a drive, browse
through the directory, and TY PR the various prospects lo find
that one file or program.

Or perhaps you're the adventurous type who thrills to the
challenge of groping through scantily labeled disks, cheering
that magical moment when hidden treasures are uncovered.

On the other hand, it occurs to us that you just may not
know the advantages of EUREKA!, the fast, menu driven disk
catalogcr for CP/M. EUREKA! puts your entire disk library
at your fingertips. Files may be found quickly and easily - by
name or by comments you can put in the file itself. Of course
the manual includes a tutorial to help you get started.

Still only SSO. Ask your dealer, or contact:

MENclociiMO Sofru/ARE Company, Inc.

Dcpt. L-l
P.O. Box 1564
Willi!!. CA 95490
Phone: (707M59.9130

Add S2.W Shipping:

Cnlil ri-vrnkiil, .llld ilOO wW iin

..i .. DlpulKcnRtCWp

CIRCLE 48 ON READER SERVICE CARD

PRODUCT BINGO

Each month Product Bingo features the latest in new soft-
ware and hardware products of interest to COMPUTER
LANGUAGE readers. Product Bingo items are based on
information received from the manufacturer and are not
meant to be product evaluations, reviews or endorsements.
To find out more about a particular product simply circle the
appropriate number on the Reader Service card — you'll
receive information directly from the manufacturer.

Note to manufacturers: Send new product information to
DougMillison, Product Bingo, COMPUTER LANGUAGE,
137 TownsendSt., San Francisco, Calif. 94107.

(8)0(8)0(8)

Application development system

Data Language Corp. has released Progress, a high-
performance application development system. In use now on
AT&T, Fortune Systems, and Convergent Technologies
machines, Progress will soon be available for the IBM PC AT
under MS-DOS and Xenix.

Progress combines a powerful data base management
system, application language, and an advanced user inter-
face. Automatic screen and report generation, error recov-
ery and an on-line tutorial are featured.

Prices start at $ 1 ,450 for single users and $ 1 ,950 for
multi-user systems. Query/run-time and plain run-time sys-
tems are available for sale with applications. A Progress
Introductory System is available for $295, including on-line
tutorial, full documentation, and all Progress facilities for
building a working application limited only by data base
size.
CIRCLE 1 1 ON READER SERVICE CARD

(x)0<8)0(x)

An APL for the IBM PC

From Portable Software comes PortaAPL, a complete imple-
mentation of the standard APL language for the IBM PC and
other systems.

Written in C for portability, PortaAPL is extended with a
full-screen editor, ASCI I character set option, Host File Sys-
tem Option, and access to machine language functions. For
faster operations, PortaAPL can work with the 8087 math
coprocessor on the I BM PC.

In addition to the IBM PC version ($ 1 ,945), PortaAPL can
be used on the Motorola 68000 ($395), and DEC VAX
($2,995). Source and OEM licenses are available.
CIRCLE 1 1 1 ON READER SERVICE CARD

®0<E>0<8>

Now your computer talks back

When the going gets tough, the tough start talking. To them-
selves. Now your computer can make it a dialogue with
RACTER, from John D. Owens & Associates. The author of
the first book written by a computer program — The police-
man's beard is half constructed — RACTER boasts the ability
to produce an original conversation on a variety of topics,

By Doug Millison

with an eccentric personality and sense of humor. Available
now for the IBM PC and CP/M, (coming soon for Apple and
Commodore), RACTER retails for $69.95.

If you want to write interactive English, checkout INRAC,
a prose-creating, high-level programming language. Com-
mands for parsing input and the ability to call subroutines at
random allow INRAC to be used to generate original prose,
avoid stereotypical responses, parody a given literary style,
or simulate specialization in a given subject. The price for
INRAC had not been set at press time.
CIRCLE 11 3 ON READER SERVICE CARD

<g)0(g>0®

New S-l 00 master, slave SBCs

Feature flexibility is the key to Teletek's new SBC 86/87 S-
100 slave SBC. Options include 1 28K to 5 1 2K RAM, Intel
8086 CPU at 5MHz or 8MHz, and a 8087 math
coprocessor.

The TurboDOS operating system provides CP/M 86 com-
patibility, while TurboDOS 1 .4 provides a PC-DOS (MS-
DOS) emulator in a multi-user environment. The SBC 86/87
sells for $837 in OEM quantities.

Systemaster II is Teletek's new S-100 master SBC. Simul-
taneous control of up to four floppy disks is possible.

Two 64K banks of parity checked RAM provide a TPA of
over 63K under TurboDOS 1 .3 and 60K under CP/M 3.0.
Two RS-232 serial ports provide synchronous or asyn-
chronous communications with software programmable
baud rates. Systemaster II sells for $679.50 in OEM
quantities.
CIRCLE 1 1 5 ON READER SERVICE CARD

Customize screen input programs

Programmers using IBM PC, AT, XT, PCjr and true com-
patibles can generate bug-free customized screen input pro-
grams quickly with Software Bottling Co.'s Screen Sculptor.

The source code generated can be modified or merged
with other programs. IBM BASIC, IBM Pascal, and Turbo
Pascal capabilities are all included in the package. Screen
Sculptor sells for $125.
CIRCLE 121 ON READER SERVICE CARD

<8>0®0<8)

Mac Switch makes sharing easy

The people who brought you Mac Inker are at it again. Mac
Switch, from Computer Friends, looks like another indispens-
able device you'll wish you had dreamed up yourself.

Mac Switch is a manual switch used to connect two periph-
erals to one computer or two computers to the same periph-
eral. It is available in two versions: parallel Centronics inter-
face for printers and serial RS-232 for data communications.
Mac Switch retails for $99.00.
CIRCLE 122 ON READER SERVICE CARD

Six Times Faster!

Super Fast Z80 Assembly Language Development Package

• Complete Zilog
Mnemonic set

• Full Macro facility

• Plain English error
messages

• One or two pass
operation

• Over 6000 lines/minute

• Supports nested
INCLUDE files

• Allows external bytes,
words, and expressions
(EXT1 * EXT2)

• Labels significant to 16
characters even on
externals (SLR Format
Only)

• Integral cross-reference

• Upper /lower case
optionally significant

Z80ASM

• Conditional assembly

• Assemble code for
execution at another
address (PHASE &
DEPHASE)

• Generates COM, HEX,
or REL files

• COM files may start at
other than 100H

• REL files may be in
Microsoft format or
SLR format

• Separate PROG, DATA
& COMMON address
spaces

• Accepts symbol defini-
tions from the console

• Flexible listing facility
includes TIME and
DATE in listing (CP/M
Plus Only)

• Links any combination
of SLR format and
Microsoft format REL
files

• One or two pass
operation allows output
files up to 64K

• Generates HEX or COM
files

• User may specify PROG,
DATA, and COMMON
loading addresses

SLRMK

• COM may start at
olher than 100H

• HEX files do not fill
empty address space.

• Generate inter-module
cross-reference and
load map

• Save symbol table to
disk in REL format for
use in overlay
generation

• Declare entry points
from console

• The FASTEST Micro-
soft Compatible Linker
available

For more information or to order, call:

1-800-833-3061

In PA, (412) 282-0864

Or write: SLR SYSTEMS
1622 North Main Street, Butler, Pennsylvania 16001

• Complete Package Includes: 280ASM, SLRNK, SLRIB

- Librarian and Manual for just $199.99. Manual only, $30.

• Most formats available for Z80 CP/M, CDOS, & TURBODOS
•Terms: add $3 shipping US. others $7. PA add 6% sales tax

CIRCLE 59 ON READER SERVICE CARD

S" L R Systems

8087 AND 80287 TECHNICAL TOOLS

87 FFT performs Forward and Inverse FFTs
on real and complex arrays which occupy up to
51 2Kbytes of RAM. Also does convolutions, auto
correlations, hamming complex vector multiplica-
tion and complex to radial conversions. Callable
from MS Fortran or 87BASIC/INLINE .... S1 50

87FFT-2 " performs two-dimensional FFTs.
Ideal for image processi ng Req uires 87 FFT...S75

MATRIXPAK" manages a MEGABYTE!

Written in assembly language our runtime
package accurately manipulates large matrices
at very fast speeds Includes matrix inversion and
the solution of simultaneous linear equations
Callable from MS Fortran 3.2, 87MACRO,
87BASIC/INLINE. and RTOS each $150

DATA ACQUISITION PACKAGE

Interactive, user-oriented language which allows
the acquisition and analysis of iarge data
streams CALL

GRAPHICS PACKAGES

Energraphics (stand alone) 295

Grafmatic for MS Fortran or Pascal 1 25

Plotmatic for Grafmatic 1 25

Halo for Basic, C or Fortran each 1 50

OTHER TOOLS

Aipha Software ESP 595

Borland Sidekick, Toolbox or Graphics 45

COSMOS Revelation 850

Lattice C 299

PSI MATHPAK 75

smARTWORK 895

SPSS/ PC 695

STSC APL# PLUS/PC 475

Micro
Way

P.O. Box 79
Kingston, Mass
02364 USA
(61 7) 746-7341

RTOS - REALTIME OPERATING SYSTEM

RTOS is a multi-user, multt- tasking real time oper-
ating system It includes a configured version of
Intefs iRMX-86. LINK-86. LOC-86, LIB-86. 0H-86,
and MicroWays B7DEBUG. Runs on the IBM-PC,
XT, PC-AT and COMPAO 4O0

INTEL COMPILERS 1

FORTRAN-86 , 750

PASCAL-86 750

PL/M-86 500

87C (LATTICE/MICROWAY) , 750

ASM-86 200

URS" • Universal Run Time System 1

Generates programs with the Intel compilers
which run on other operating systems MS-DOS
version is included with RTOS
Xenix-286 Version 300

SoftScope Symbolic Debugger 1 soo

'Requires RTOS or iRMX-86. All Intel compiler
names and iRMX-86 TM Intel Corp

87 BASIC/INLINE™ converts the output of
the ISM Basic Compiler into optimized 8087
inline code which executes up to seven times
faster than 87BASIC, Supports separately com-
piled inline subroutines which are located in their
own segments and can contai n u p to 64 K bytes of
code This allows programs greater than 128K!
Requires the IBM Basic Compiler and Macro
Assembler. I ncludes 87 BAS IC $200

PC AT and 86-31 DRIVES

30 MEGABYTE WINCHESTER 2000

53 MEGABYTE WINCHESTER 2600

SYQUEST FIVE MEGABYTE 950

VbuCan
TalkToUs!

HARDWARE AND LANGUAGES

8087-3 5mhz $149

Including DIAGNOSTICS and 180- day warranty
For IBM PC and compatibles

8087-2 8mhz $275

For Wang ATST. DeskPro, NEC. Leading Edge

80287-3 5mhz $275

For the IBM PC AT

64K RAM Set $30

256KRAMSet $195

128K RAM Set pc at $225
NUMBER SMASHER™ call

1 0mhz 8087 coprocessor board for the IBM PC
FORTRAN and UTILITIES

Microsoft Fortran 3.2 239

IBM Professional Fortran 595

Intel Fortran-86' 750

FORLIB+ 65

STRINGS and THINGS 65

BASIC and UTILITIES

IBM Basic Compiler 270

87BASIC/INLINE 200

Summit BetterBASIC" 1 75

Summit 8087 Module 87

MACRO ASSEMBLERS

IBM Assembler with Librarian 1 55

87MACRO 1 50

PASCAL

Microsoft Pascal 3.2 209

Borland Turbo. 45

Turbo with 8087 Support 85

NO CHARGE FOR CREDIT CARDS

ALL ITEMS IN STOCK
CALL FOR COMPLETE CATALOG

CIRCLE 57 ON READER SERVICE CARD

72 COMPUTER LANGUAGE ■ FEBRUARY1985

SOFTWARE REVIEWS

Expert team analyzes 21 C compilers

By Steve Leibson, Fred Pf abler, Jim Reed, and Jim Kyle

n the early 1970s at
Bell Laboratories,
: Dennis Ritchie
developed his C programming language
for PDP-1 1 minicomputers. He needed a
systems programming language for the
new UNIX operating system created by
colleague Ken Thompson. The C lan-
guage has grown in popularity since those
early days due to its simplicity, elegance,
and power.

Today C has left the UNIX environment
and comfortably resides on other oper-
ating systems and computers. This special
review will look at the C compilers cur-
rently available on the market from a vari-
ety of operating system perspectives. We
will cover 12 C compilers and one inter-
preter running under PC-DOS, six CP/M
C compilers, two OS-9 C compilers, one
CP/M 86 C compiler, and one TRS-80
compiler. (C compilers for the Apple
Macintosh will be reviewed in the April
issue.) A list of company names,
addresses, and phone numbers is included
at the end of the review.

Languages have been grouped by oper-
ating system in the text and in four tables
on bundled software and essential infor-
mation (page 75), benchmark results
(page 79), basic data types (page 99), and
library functions (page 100). It may not be
fair to compare language implementations
across operating system boundaries, but
such comparisons are inevitable and
interesting.

We tested the C languages by installing
them in appropriate systems. Then we
attempted to compile and run a suite of
four benchmark programs on each.
Details on the benchmarks we used-
Sieve of Eratosthenes, Fib, Deref, and
Matrix—arc contained in a sidebar
(page 82).

Throughout this review we refer to the
definitive reference on C — The CPro-
grammirtg Language by Brian W. Ker-
nighan and Dennis M. Ritchie. C users all
over refer to this book simply as K&R and
we will too.

MS-DOS/PC-DOS C COMPILERS

Little doubt exists that MS-DOS, in the
guise of PC-DOS in the IBM PC family
and the many compatibles, is the succes-
sor to CP/M as the most popular micro-
computer operating system. With the
addition of tree-structured directories
introduced with DOS 2.0, this operating
system clearly targets low-end UNIX
applications.

In this atmosphere, many software
houses have decided to offer C compilers
for MS-DOS and PC-DOS. As a bonus,
one company has introduced a C inter-
preter! We will look at 1 1 of these com-
pilers and the interpreter in this portion of
the review. All compilers were tested
using DOS 3 .0 on an IBM PC with 544K
of memory and two 360K floppy drives.

The 8088 processor, with its segmented
architecture, presents an interesting prob-
lem for language developers. Pointers can
be either 16 bits or 20 bits long. This has
led to the use of four programming mod-
els, originally set forth in Intel's PL/M 86
programming language. These models are
called small, medium, compact, and large
(or big).

The small 8088 memory model defines
separate code and data areas, with each
area limited to 64K bytes. Thus the total
program with data included may use a
maximum of 128K bytes of memory. The
program stack must also fit in the 64K
allocated to the data. Programs compiled
with the small memory model use 16-bit
pointers.

The medium memory model limits the
data segment to 64K bytes, so data point-
ers within the program are 16 bits long.
The program, however, is no longer con-
strained to a 64K block but can extend
throughout memory. Call and return
pointers are therefore managed by the
compiler and linker so that their size is
appropriate for each instance.

Programs compiled under the compact
memory model have separate 64K spaces
for code, data, and stack segments, while
a fourth segment, called the heap, can
occupy the rest of memory.

The large or big model allows code to
expand throughout memory, while the
global data and stack are limited to 64K
segments. The heap can also extend

throughout memory, as in the compact
model.

Lattice and Microsoft have a slightly
different set of memory models; the
small, program, data and large models
allow for a simpler allocation of pointer
size. Small programs have a 64K code and
a 64K data segment. Programs compiled
with the program model have a 1 MB pro-
gram space but a 64K data space, while
programs compiled with the data model
are limited to a 64K program space and
have a 1 MB data space. The large model
allows for both segments to use 1MB.

Manx's Aztec C

Let's start off with a fine example of a
compiler that has many strengths— one
from Manx Software Systems with the
unlikely name of Aztec C. Two versions
of the package are supplied, version
1 .06D and 2.20B. The earlier version is
slower but more mature (less bugs), and
the new version is faster but has not yet
aged. We tried both versions, and the
results arc in the benchmark table (Table
2). Programs compiled with the version
1 .06 compiler won't link with the version
2.20 linker and vice versa. You get a "not
an object file" error.

Two floating point libraries arc sup-
plied. One library does not use the 8087
and the other does. Aztec C supports only
the small memory model, but it also sup-
ports overlays, so programs can become
quite large. The version 2.20 compiler
has an option to support either the
8086/8088 or the 801 86/80286. We did
not quite understand this because the
80 1 86 is more like the 8086 than the
80286, however, we did not have the hard-
ware to test this option.

No batch files were supplied on the dis-
tribution disks and we found no recom-
mended files in the manual, but the syntax
for the compile and link commands is
fairly simple. With the compile disk on
drive A and the source code on drive B,
CCB: < filename > compiles the pro-
gram and LN B: <filename > CUB links
it.

The Sieve and Fib programs compiled,
linked, and ran for both versions of Aztec
C, but we had problems with Dercf and

Of course,
POWER! saves
your Bad Disk.

NOW! WINDOWS FOR IBM!

Italsoaoes "r^r *$*
54 other things to Hp^
keep your disk in line.

EVERYTHING YOU ALWAYS WANTED
TO DO, BUT WERE AFRAID TO TRY

Unlike some utility programs that are a headache
to use, POWER! is engineered to spoil you with 55
features, simple and uniform commands, and utter
simplicity of use. POWER! automatically alpha-
betizes and numbers your files. You select by the
number and never tvpe tile names again. Need to
[COPY]. [RENAME], [ERASE], or [RUN] programs? Just
type in their menu number! POWER! also locks
out your disk's bad sectors [TEST] without destroy-
ing files-a critical difference from other utilities
that search and destroy, without informing you
what they've done, leaving you to wonder why
your programs won't run. (And POWER! still has
50 commands to go!)

POWER! ONE PROGRAM DOES IT ALL!

You may own a few utility programs for your com-
puter housekeeping, each with its own commands
to memorize. POWER! has all the programs rolled
into one 16K integrated package, so you do things
you've never tried before-every day. Save sen-
sitive data from prying eyes with [PASS] word pro-
tect, move a block of memory [MOVE], look fur data
[SEARCH] or compare tiles [CHECK]. POWER! also
makes easy work of patching, [DISPLAY/SUBSTITUTE],
customizing software [LOAD/SAVE]. Among the
other commands are [SIZE], [STATj [LOG], [DUMP].
[TYPE]. DUMP], [Fill], [SET], and the CP/M version
lets you restore erased files-even when you don't
remember the lilename-at a Hick of the POWER!
[RECLAIM] command, (Still !il commands to go!)

POWER! NOW FOR IBM's PC-DOS
AS WELL AS CP/M

We first developed POWER! for CP/M two years
ago, and a stack of testimonials from FORD to
XEROX testify to its excellence. For IBM-PC"
users, special features like managing sub-direc-
tories, [CHANGE], and a separate creation of up to
8 simultaneous, on-screen [WINDOWS] have been
added.

MONEY-BACK GUARANTEE AND
A 10 DAY TRIAL

POWER! has the Seal of Approval from the Pro-
fessional Software Programmers Association, and
you, too, must be happy with POWER! -or your
money back ! For only $169 you can now really be in
control of your computer. Call Computing! at (415)
567-1634, or your local dealer. For IBM-PC or any
CP/M machine. Please specify disk format.

, ■ I <|

Jl'llii!

The company that earns its exclamation point.

COMPUTING! _

2519M Greenwich, San Francisco, CA 94123

TO ORDER CALL 800 TOLLFREE

800-428-7825 Extension 96M
In CA: 800-428-7824 Extension 96M

MM and IBM PC are rmrfmd trademarks of
intimations! B usine ss Macaittn Corporation.

CIRCLE 37 ON READER SERVICE CARD

74 COMPUTER LANGUAGE ■ FEBRUARY1985

Matrix. Deref had to be reduced to 12 lev-
els of indirection before Aztec C could
compile it. At that level, the program
compiled and linked in 60 sec, ran in 6
sec, and took 4,352 bytes.

Our problem with Matrix turned out to
be library related. A second math library,
called m.lib, must also be linked with
Matrix. After that, we had no further
trouble.

Manx offers Aztec in several different
flavors, so you can port your programs
across a lot of hardware if you need to.
One interesting aspect of the Aztec C
compiler is the PRO extension. This
includes an editor (called z), a symbolic
debugger, and source files for several of
the library functions. The PRO version of
the compiler can generate ROMable code.
These extensions seem to be quite valu-
able for serious development work. The
Aztec C86 compiler for MS-DOS or
CP/M 86 is $249.00, and the compiler
with the PRO options is $499.00,

Supersoft C

Supersoft C went together a little differ-
ently from the other packages. It also
would not fit on one 360K disk so we were
instructed to put just the library on a sec-
ond disk. Then a single batch file called
CM. BAT ran the show.

When it came time to link in the library,
the linker paused and asked where the
library could be found since it wasn't
where it was supposed to be. This was our
cue to swap the library into drive A and
type "A". Well, that didn't work but typing
"A:" did. Another company that doesn't
seem to read its own manual .

At $350, Supersoft C seems a bit
expensive for a compiler that does not yet
support floating point. Supersoft says it
plans to upgrade the compiler to floating
point but for now it supports Binary
Coded Decimal floating point through
float and double declarations, plus some
special library math routines. It seems
that although you can declare floats and
doubles, all you can do with them is pass
their pointers around. You can't use the
compiler's intrinsic operators on them.

Only one memory model seems to be
supported by Supersoft C, and it appears
to be the small model. Supersoft C does
not directly support overlays, but a sec-
tion in the manual states that Digital
Research's RASM86 assembler running
under MS-DOS will support overlay oper-
ation. Supersoft's linker ELINK supports
overlays in the CP/M 80 environment.

In the benchmarks. Sieve compiled and
ran fine, we had to remove void from the
Fib program, Deref ran fine, and Matrix
could not be compiled due to the lack of
intrinsic floating point math.

Supersoft C seems to be a software
package in transition. Until it supports
full C, wc think it is overpriced. When
floating point math is added, it will be a
real contender.

Computer Innovations' C86

We liked Computer Innovations' C86
right away. You have to admire a company
that names its librarian Marion. However,
this fondness quickly dimmed . Exe-
cutable files in C86 are supplied squeezed
to save room on the disk and must be
unsqueezed to create the working disk.
We don't know about you, but wc like to
get the software out of the box and get it
running fast. Unsqueezing files is for bul-
letin boards!

Learning how to use C86 isn't as easy
as it might be. No test program is sup-
plied, although there is a listing in the
manual. One paragraph is devoted to link-
ing. We think that this is hardly adequate.
In the end we stole the link batch file
LNK.BAT from the C-Systems work
disk.

C86 supports the small and big models.
The 8087 is also supported by a compiler
option and a special 8087 library. Pro-
grams compiled with this option selected
will not run on machines that do not have
the 8087 coprocessor. Computer Inno-
vations previously supported overlays,
but with the addition of support for the big
memory model, overlays were dropped.

Sieve compiled and ran fine but Fib
wouldn't parse. It was then that we found
out that the C86 manual lists no error
messages. It only says that error messages
"should be self-explanatory." We
wouldn't have looked in the manual if that
were true, honest. The problem with Fib
was the void typing again. Removal of the
offending void allowed Fib to compile and
run.

Deref proved to be too complex again.
Reducing the indirection to 14 levels
proved to work. At that level of complex-
ity, Deref compiled in 106 sec, ran in 6
sec, and took 12,680 bytes. Matrix
required a ttinchtde <sidio.h > before it
would compile.
Wizard C

There is nothing like a good name to get a
product off the ground quickly, and we
like the name Wizard. It makes us feel
like real hackers. Wizard Systems' Wiz-
ard C seems to have been written by
someone who has been in the business a
while. This especially shows in the docu-
mentation, which is conversational with-
out being preachy or snobbish.

Wizard C barely fits on a double-sided
floppy. We were finally able to squeeze all
of the required files onto the disk by delet-
ing COMMAND.COM. This makes the
disk unbootable but much easier to use
during compilation.

Using a confusing set of compiler
options, you can generate programs under
the small, medium, or large model. The
default is the small memory model.
Adding the -e option gives you extended
data, but this option must be used with the

Bundled software and essentia! information

Lattice

$500

2.13(2.14
avail)

MS-DOS,
CP/M 86

N.A.

yes

OBJ

yes

150 +

Microsoft

S500

2.03

MS-DOS

N.A.

yes

OBJ

yes

150 +

Microsoft

S450-500

3.00.13

MS-DOS

N,A.

yes

OBJ

yes

832

Mark
Williams

MWC86

SSOO

2.0

MS-DOS,
Coherent,
CP/M 86,
ISIS, RMX,
VAX/VMS

yes

OBJ

yes

238

Rational

Instant C

S495

.9(1.0

MS-DOS

yes

N.A.

yes

N.A.

yes

175

„ Systems

avail]

'5. Wizard
| Systems

Wizard C

S450

l.3|2.0
avail)

MS-DOS,

TOS,

VAX/VMS

OBJ or

ASM

yes

243

£1 Computer
^ Innovations

Optimizing
C86

S395

2.20G
(2.20J

MS-DOS,
CP/M 86

yes

OBJ or

ASM

yes

275 +

• Supersofl

S350

avail)
1.2

MS-DOS,
CP/M,
CP/M 86

yes

ASM

228

Manx
Software

Aztec O
PRO

S498

2.20B

MS-DOS,
CP/M 86,
Macintosh

yes

ASM

yes

400 +

C-5ystems

SI99

2.06(2.08
avail 2/1)

MS-DOS

yes

ASM

yes

193

CWare

DeSmet C

$159

2.4

MS-DOS,
CP/M 86

yes

128

Datolighl

Small-C

S40

2.1

MS-DOS

S40

ASM

yes

Ecasoft

Eco-C

S250

3.1(3.2
avail 1/1)

CP/M

yes

ASM

yes

210

Alcor

Alcor C

$139

2.01.00

CP/M,
MS-DOS

yes

p-code

yes

486

j> BDS
'5.

BDSC

$150

1.50a

CP/M,
MP/M

N.A.

yes(2)

yes

CRL

yes

184

| Code Works

QIC

$95.00

3.2B[3.2C
now avail)

CP/M,
MP/M

ASM

yes

173

Software
Toolworks

C/80

$49.95

3.1

CP/M,
MP/M,
MS-DOS

yes

ASM

yes

Hendrix

Small-C

$25.00

2.1

CP/M,
MP/M

ASM

yes

256

£ Microware

$400

1.1.4

OS-9

yes

ASM

yes

ASM

yes

132

13.

v Inlrol
CJ
(/)
O

$425

1.55

OS-9,
Flex,

Uniflex

yes

level
no

yes

ASM

yes

218

m Digital

^ Research

$500

1.11

CP/M 86
MS-DOS

yes

OBJ

yes

500 +

N.A. = Not

Applicable.

Table 1.

,<*&«*

^SSfLeS

,<*>*

.Kp<>»

6^" \»^„#'

■.^^rV

1£^

^ V*

«*&

>£<*

MICROTEC
RESEARCH

3930 Freedom Qicie Suite W . Santa Cuts CA %05J
Mailing Address. PO Bo> 60337. Sunnyvale. CA §4088
14081 733:2519 • Tele* 'IT!) 4990808

CIRCLE 71 ON READER SERVICE CARD

-m option to create a large model pro-
gram. The -m option used alone generates
a medium model program. A nice feature
of the compiler is that it automatically
invokes the linker, reducing the need to
babysit the compilation process.

One very useful program in the UNIX
C programming environment is LINT.
This is not something you pick off of your
clothes but an extended error checking
program designed to aid in writing code.
Wizard C'has LINT-type error checking
built into the compiler. This feature pro-
duced several interesting warnings during
compilation of the benchmark programs.
In addition. Wizard C is supplied with the
LINT program, and phase one of the com-
piler can be told to generate lint files
(.LNT) to be used by LINT.

The Sieve ;md Fib benchmarks com-
piled and ran fine, but Dercf struck once
again. Wizard C came up with the very
descriptive message "Disaster, Type stack
overflow." The compiler could only han-
dle 16 levels of indirection before the
stack blew. At that level of indirection,
Deref compiled in 72 sec. ran in 8 sec,
and required 8,444 bytes of code. Matrix
required a ^include <sidio.h > before it
would compile.

Datalight's Small-C
As an indication of MS-DOS's lineage,
we find the migrant Small-C fresh from
its appearance in the CP/M world. This is
J. E. Hendrix's version 2. 1 of Small-C,
retargeted for the 8086 processor family
by Datalight. Most of the words written
later on about the CP/M version of Small-
C apply to the MS-DOS version as well.
Smatl-C for MS-DOS supports only the
small memory model,

Small-C has a loyal group of followers,
which is most evident from the number of
programs written in the language appear-
ing in Dr. Dobb 's Journal over the past
few years. Therefore, we find this version
for PC-DOS and MS-DOS to be most
welcome.

Small-C requires but is not supplied
with an assembler and a linker. We used
the Microsoft macro-assembler MASM
version 1 .0 and the linker supplied with
DOS 3.0 in our tests. Two programs sup-
plied with Small-C were batch file
CC.BAT, which can be used to automate
the compilation process, and LL.BAT,
which docs the same for linking. CC.BAT
worked fine but LL.BAT would not work
with version 2.2 of the DOS linker as sup-
plied. This was due to the specification of
the bit bucket NUL: as the target for the
linker's list file. Apparently LINK ver-
sion 2.2 (supplied with DOS 3.0) prefers
NUL (without the :). A quick swipe of the
editor and Small-C was off and running.

In the benchmark tests, Small-C com-
piled and ran Sieve with no problem but
hit the rocks on the other three programs.
It couldn't compile Fib until references to

void and unsigned were removed. Then it
was able to compile and properly run the
benchmark. Deref requires structures and
Matrix requires floating point arithmetic
and two-dimensional arrays. Small-C
does not support these three features so
the last two benchmarks weren't compiled
or run.

At $40, Small-C really isn't intended to
be a heavy-duty development tool
although it can be remarkably useful. It
serves as a low-cost introduction to the C
programming language and an instructive
example of how to write a compiler.

Small-C is the only MS-DOS C compiler
we reviewed here which includes the
source. The library source is also
included. Thus you can learn how com-
pilers are constructed and modify the
existing Small-C compiler
experimentally.

C Ware's DeSmetC

One rung up the price ladder from Small-
C is a terrific package from C Ware called
DeSmetC. For $159 you get a load of

Btrieve and Xtrieve.

B-tree file access for all your programming

languages. With the query tool your users demand.

Introducing a powerful data base com-
bination for PC application develop-
ers. Btrieve"'', the most sophisticated
file access method for your IBM™ PC.
In single user and network versions.
And Xtrieve™, a new menu-driven
query tool that gives you— and your
users— fast access to information.

Btrieve: for professional
programmers. Btrieve provides fast,
flexible file management for all your
application development. All your pro-
gramming languages — BASIC, Pascal,
Cobol, C. With multikey access to re-
cords. Automatic file recovery. Unlimit-
ed records per file. Duplicate, modifiable,
and segmented keys.

With Btrieve, you can develop better
applications faster.

Xtrieve: easy window interface.
The ideal complement to Btrieve,

Xtrieve is the non-programmer's inter-
face. Xtrieve's full relational capabil-
ities let users define a virtual table of
data from multiple files. Then Xtrieve
speeds them through query building
with a series of easy-to-follow windows.

No command language. Xtrieve is
completely menu-driven, so there's no
need to memorize command language.
Or special syntax. Everything you need
is on the screen.

Xtrieve features a full range of restrict-
ion criteria. Online help messages. And
interfaces to access information from
Lotus 1-2-3™ and dBase II® files.

Network capabilities. Network ver-
sions of Btrieve and Xtrieve allow data
sharing in PCnet™, NetWare™, Ether-
Series™, MultiLink™, and OmniNet™
LANs.

For more information or to order,
contact:

SoftCraftlnc.

P.O. Box 9802 #590 Austin, Texas 78766 (512)346-8380

Suggested retail prices: Btrieve, $245; Xtrieve,
$195. Btrieve/N (network), $595; Xtrieve/N,
$395. Dealer inquiries welcome. Btrieve
requires PC-DOS or MS™-DOS l.X or 2.X;
Xtrieve, PC-DOS or MS-DOS 2.X.

Btrieve and Xtrieve, IBM, 1-2-3, dBase 11, PCnet,
NetWare, EtherSeries, MultiLink, OmniNet, and MS
are trademarks of SoftCtaft Inc., International Business
Machines, Lotus Development Corp., Ashcon-Tate,
Orchid Technology, Novell Data Systems, 3Com Corp.,
Davong Systems Inc,Corvus Systems, and MicioSoft Inc.

CIRCLE 78 ON READER SERVICE CARD

ORDER BACK
ISSUES OF

COMPUTER
LANGUAGE!

Premier .
Oct. '84 .
Nov. '84.
Dec. '84 .
Jan. '85 .
Feb. '85 .

NAME_

. copies x $4.00 = $_
. copies x $4.00 = S_
.copiesx$4.00=$.
. copies x $4.00 = $-

copiesx$4.00 = $.

copiesx$4,00 = $.

COMPANY.
ADDRESS_

CiTY, STATE, ZIP

Send payment and
coupon to:

COMPUTER LANGUAGE

Back Issues

131 Townsend St.

San Francisco, CA 94107

WRITE

The Writer's Really Incredible Text Editor lives up to its
name! It's designed for creative and report writing and
carefully protects your text. Includes many features
missing from WordStar, such as sorted directory listings,
fast scrolling, and trial printing to the screen. All editing
commands are single-letter and easily changed. Detailed
manual included. Dealer inquiries invited. WRITE is
$239.00.

BDS's C Compiler

This is the compiler you need for learning the C language
and for >writing utilities and programs of all sizes and
complexities. We offer version 1.5a. which comes with a
symbolic debugger and example programs. Our price is
(postpaid) SI 30.00.

Tandon Spare Parts Kits

One door latch included, only $32.50.

With two door latches S37.50.

Door latches sold separately for S7.00.

All US orders are postpaid. We ship from stock on many
formats, including: 8", Apple. Osborne. KayPro, Otrona.
Epson, Morrow, Lobo, Zenith. Xerox. Please request our
new catalog. We welcome COD orders.

Workman & Associates

112 Marion Avenue
Pasadena, CA 91 106
(818) 796-4401

«OLV * U

CIRCLE 68 ON READER SERVICE CARD

*UNIX System III POWER and sophistication are yours.

Let THE SOLUTION turn your micro into all you

dreamed it could be, bringing the Ultimate

programming environment as close as

your modem. Just a local call

from over 300 cities

nationwide via Telenet

EXPANSIVE SOFTWARE DEVELOPMENT FACILITIES including Language and Operating System design.
• LANGUAGES: C, Fortran 77, RATFOR, COBOL, SNOBOL, BS, Assembler + Artificial Intelligence
programming via LISP,

• USENET Bulletin Board System— 800 + international UNIX sites feeding over 190 categories,
typically bringing you more than 160 new articles per day.
■l. • Interuser and Intersystem mail + 'chat' capability.
P||, • UNIFY: Sophisticated data-base management system.

^j V • UNIX & System enhancements from U.C. Berkeley and Korsmeyer Electronic Design Inc.
. g? * Online UNIX manuals + Expert consultation available.

^ , , • SOLUTION-MART: Hardware/Software discount shopping database.
• LOW COST and FAST response time,

(as low as $8.95 hr. connect time + $.05 cpu sec. non-prime)
• $24.95 = 1 hr. FREE system time + SOLUTION News subscription + BYTE
BOOK (Introducing The UNIX System 556 pp.).

—■ * UNIX is a trademark o! Bell Labs.

Payment via VISA or Master Card

— ELECTRONIC DESIGN, INC -

CIRCLE 34 ON READER SERVICE CARD

5701 Prescott Avenue
Lincoln, NE 68506-5155
402/483-2238
10a-7p Central

Benchmark results

Compiler

Lattice

Compile

time
(see)

-Sieve —
Run
time
(sec)

Size

(bytes)

(sec)

Compile
time

(see)

-Fib-
Run
time
(sec)

Size

(bytes)

(see)

13,710

Compile

time

(see)

-Deref-
Run
time

(see)

Size
(bytes)

Compile
time

- Matrix -
Run
time

Size
(bytes)

21,902

13,726

163

25,176

Microsoft
(2.03)

103

19,754

11,578

22,722

Microsoft
(3.0)

102

6,926

109

6,924

104

6,940

—

Mark
Williams

6,887

—

6,869

6,874

107

10,847

Rational
Systems

640

176

640

—

1,408

O
Q
O

-*,
(A

Wizard
Systems

16,490

8,444

—

18,826

Computer

Innovations

111

12,729

112

12,728

—

134

13,766

a
in

Supersoft

132

13,112

114

4,738

119

4,822

Manx 1.06D

4,448

4,384

7,920

Manx 2.20B

4,448

7,840

C-Systems

119

27,716

112

19,456

191

31,114

CWare

6,656

7,168

Dalalight

14,182

5,861

Digital
Research

108

27,136

109

1 8,944

145

18,944

145

31,872

Ecosoft

14,464

154

Alcor

215

1,024

'5.
E

BOS

3,840

113

3,840

3,712

Code Works

12,160

104

4,352

Software
Toolworks

117

12,288

114

3,584

—

172

12,288

Hendrix

270

16,640

Microware
{from disk)

111

3,873

Ill

3,812

—

151

8,207

'5-
£

Microware
(from RAM]

3,873

3,812

—

112

8,207

Introl
{from disk)

195

7,618

195

7,546

191

7,567

246

8,370

Introl
(from RAM]

—

7,618

—

7,546

—

7,567

—

8,370

_o
'a.
E

Digital

Research

(CP/M-86)

158

28,672

155

20,480

153

20,480

195

34,816

Digital

Research

[MS-DOS)

108

27,136

109

18,944

145

18,944

145

31,872

Table 2.

EVERY PROGRAMMER/ANALYST NEEDS . . .

BEMffiM

THE STRUCTURED PROGRAMMING TOOL!

40*

1. DESIGN

PROGRAMS

2. CODE
PROGRAMS

3. DOCUMENT
PROGRAMS

BUILD DATA

STRUCTURE
DIAGRAMS

BUILD PROG
STRUCTURE
DIAGRAMS

ASSIGN

EXECUTABLE
OPERATIONS

GENERATE

SOURCE

CODE

"FILE"

DIAGRAMS
ON DISK

"LIST"
DIAGR
ON PR

AMS

INTER

1 O O i O

O | O O I O

ADA

DBASE II
COMMAND

FILES

PASCAL

PL/I

STRUCTURED
BASIC

SCHEMATIC
LOGIC

Design your programs right on the screen using modern techniques based on the popular Jackson
Structured Programming method (JSP)!

DEZIGN is more than just another flowcharting tool. It is an integrated tool for designing and
documenting programs and for generating source code in any of the languages listed above!

DEZIGN enables a software developer to create Data and Program Structure Diagrams using the
Sequence, Selection, and Iteration constructs; assign detailed statements to the Diagrams; and
synthesize source code from the control logic represented on the Diagrams and the detailed
statements assigned to them.

DEZIGN lists for $200 and is available directly from Zeducomp, from IBM and Zenith personal
computer dealers, and from software retailers.

• DEZIGN-PC runs under DOS 2.0 and 2.1.

• DEZlGN-86 runs under CP/M-86 1.1 and Concurrent CP/M-86.

• DEZIGN-1 6 runs under ZDOS on Zenith Z-1 00 computers.

"Principles of Program Design," by M. Jackson, Academic Press, describes the JSP metholology
and is available for$50 directly from Zeducomp and from most professional/technical bookstores.

P.O. BOX 68 • STIRLING, NJ 07980 • (201) 755-2262

dBASE II is a trademark of Ashton-Tate, Inc.
CIRCLE 91 ON READER SERVICE CARD

software — a complete development .sys-
tem, in fact. You start with a screen editor
geared to the IBM PC. You can pay well
over S159 for just an editor and we have
just begun.

Next is the compiler. It isn't a small
compiler at all. but a nearly full K&R
implementation. What's missing? Well,
some of the data types, preprocessor com-
mands, storage classes and an operator or
two. But these omissions didn't stop
DeSmct C from compiling and running
the four benchmarks with no
modifications!

The compiler supports only the small
memory model but makes up for the
memory limitations by supporting over-
lays. Support is also available for the Intel
8087 numeric coprocessor. Two libraries
are supplied with DeSmct C. One sup-
ports floating point calculations with
assembly language routines and the other
uses the 8087.

Also included in this software bargain
bonanza are a binder (linker), an assem-
bler, a performance profiler (!) and a
librarian. The linker is necessary because
the DcSmet compiler output is not the
.OBJ file the DOS linker can use. The
assembler is not needed for the com-
pilation of C programs. It is a parallel
development tool for routines that must be
coded in assembly language for effi-
ciency. The assembler produces object
files compatible with the DcSmet binder.
Do you know what a profiler is? It is a
jolly little utility that uses the IBM PC's
real-time clock to generate a histogram of
your program's operation. It docs this by
latching on to a periodic interrupt and
sampling the instruction pointer (IP) at
each interrupt. The value in the IP is used
to increment one of ! ,024 counters or
"buckets". The result is a good approxi-
mation of where your program spends its
time. Current versions of DcSmet C allow
the profiler to work symbolically.

We did not have a chance to work with
the DcSmet debugger or librarian, but
they would appear to be most adequate
from the documentation. Speaking of the
documentation, we get to our only com-
plaint about DcSmet C. For SI 59, we
think C Ware could afford a vinyl 3-ring
binder. The documentation is good, but it
needs packaging.

Finally, DeSmet C was the easiest DOS
C compiler to bring up. Twenty minutes
out of the box, it compiled a program.
That includes the time it took to create
working copies of the compiler and
binder. It worked just like the instructions
said it would too! Truly a miraculous
package.

C-Systems C

Returning to earth, we look at C-Systems
C. This compiler is more a complete
implementation of K&R although still not
a full version. C-Systems C docs not come

with an assembler or linker. Using the
DOS assembler and linker, the complete
development package of compiler, assem-
bler, linker and library won't fit on one
360K floppy disk. Thus you will need a
compile disk and a link disk and will be
swapping between them unless you have a
hard disk or RAM disk.

The compiler supplied for $199.00 sup-
ports the small and big memory models.
Libraries for the compact and medium
models are S50.00 each. Library source is
not included but is available at an extra
charge. The 8087 is also supported. It can
be sensed automatically or the compiler
can be instructed to assume that an 8087
will be present, for maximum
performance.

This compiler did not work the way the
manual said it would. A batch file called
CC.BAT is supplied to run the com-
pilation. You are supposed to have your
source on drive B and the compile disk on
drive A. Then, while logged on to B. type
"a:cc <filename>" and your file will
compile. Except it won't. CC.BAT
expects the compiler files to be on the
default drive, which in this case is B, but
the files are actually on A. Logging onto
drive A and typing "CC B: < filename >"
works ever so much better. You'd think
they would have tried it, wouldn't you? A
batch file for linking, called LNK.BAT.
did work as advertised.

Another problem we had in running C-
Systems C the first time was due to our
own error, with plenty of help from C-
Systcms. They supply batch file CC.BAT
and an executable file called CC.EXE on
the same disk, and we copied them both
over to our working disk without looking.
CC.EXE doesn't perform the same as
CC.BAT, yet it overrides it when CC is
called from DOS. We shouldn't have
copied CC.EXE to our working disk and
neither should you unless you know what
it does. Our opinion: C-Systems should
not supply its programs with this potential
for error built in.

Though apparently a more complete
implementation of C than DeSmct, C-
Systems C had problems in the bench-
marks. Sieve compiled and ran fine, but
problems cropped up getting the other
three to compile. C-Systems C didn't like
void in Fib, couldn't handle 40 levels of
indirection in Dcref. and required a
tfinclude <sidio.li > in Matrix.

Since the compiler couldn't deal with
the indirection in Deref, we started reduc-
ing the level of complexity until the pro-
gram would compile. The magic number
was 30 levels of indirection. No values
were entered in the benchmark table for
this program since it is not the full pro-
gram, but at 30 levels of indirection,
Deref compiled in 106 sec. ran in 9 sec,
and linked to 19.426 bytes.

Easy to Use

Word Processor with the

Power of BASIC

VEDIT
has been

acclaimed
for the last
four years
the industry
standard
in text
editing.

p «-forin anc ,,
BweofUe

5 -i ? ?
£ I I J

°ao 01

° D q
c D D 0.

Now there's VEDIT PLUS.

VEDIT PLUS is easy to learn, yet
can do things far beyond ordinary
word processors:

• Sort a mailing list

• Perform arithmetic computations

• Compare files

• Edit multiple files of unlimited size

The power of VEDIT PLUS is equal to
a high-level language - it gives you:

• If-then-else decision making

• Pattern matching

• User prompts and input

• An optional Z80 to 8086 translator

Expect a lot from VEDIT PLUS. It's
from CompuView - nationally recog-
nized for user support.

For a demo disk, more infor-
mation or the dealer nearest you call
toll free:

1-800-327-5895

VEDIT PLUS is available for prac-
tically every microcomputer . . .$225

CompuView

PRODUCTS, INC.

1955 Pauline, Ann Arbor, Mi 48103
(313) 996-1299 • (800) 327-5895

CIRCLE 40 ON READER SERVICE CARD

Benchmarks

The suite of four benchmark programs
we used in reviewing the C languages
was taken from a recent article in
BYTE magazine. 1 Written by Dr.
David D. Clark, these benchmarks test
out various features of the languages.

First, however, we found it neces-
sary to debug the benchmarks. Some-
thing must have happened to Clark's
programs between benchmark testing
and publication because two of the
programs listed wouldn't run. If any-
body would like a copy of our
reworked benchmarks, written in C,
they are all available on the
COMPUTER LANGUAGE Bulletin
Board Service and on this magazine's
account on CompuServe. Look for the
files BENMRK. *. With all the tables
and lengthy text from this review,
there was just not enough room in the
magazine left to print these
benchmarks.

The first benchmark we used is the
ever popular Sieve of Eratosthenes,
written in a very simple dialect of C.
This program is also called the prime
number generator. It is a good test of a
language's ability to run loops since
the program has a triply nested loop.
Sieve computes all prime numbers
from 3 to 16,38 1 • It does this by
manipulating an array of 8,191 flags,
with each flag representing an odd
number from 1 to 16,381, Even num-
bers are obviously not prime and thus
require no flags in the array. At the
completion of the program, the largest
prime found (16,381) is printed as is
the number of primes ( 1 , 899) . All
compilers were able to run this pro-
gram and obtain the correct answer.

The second benchmark, called Fib,
generates Fibonacci numbers. This is a
recursive algorithm so the program
tests the speed of function calls and the

ability of the language implementation
to perform recursion. The Fibonacci
numbers are computed using the fol-
lowing algorithm:

lfx<=2rhenF(x) = 1
lfx>2thenF(x) = F(x-l) +
F(x-2)

When the program completes, it prints
the value of F(24) , which is 46,368.

Though our third benchmark is a very
short program, many of the compilers
had a lot of trouble with it. Called
Deref, it tests the speed at which a C
deals with indirection. By setting up a
nested structure, the program repeat-
edly accesses a variable through 40
levels of indirection. This complexity
of indirection was managed by only a
few compilers.

The fourth benchmark was derived
from a program proposed by BYTE
columnist Jerry Pournelle. 2 Called
Matrix, it tests floating point arith-
metic speed. Since some of the com-
pilers reviewed in this article do not
implement floating point arithmetic,
they weren't able to compile this
benchmark. C programs don't always
require floatingpoint calculations so
the inability to run this benchmark is
not a terrible blot on a compiler's
record.

Still, it is valuable to compare floating
point performances on those C imple-
mentations that can run this program.
The program creates three matrices. It
then fills (initializes) the matrices and
computes values by iteratively per-
forming matrix multiplication and
addition. This program is designed to
be very compute-intensive. The final
value printed is the sum of all elements
in the C matrix, which is 465,880.

Rational Systems' Instant C

And now for something completely differ-
ent, as Monty Python would sav. Instant C
from Rational Systems is a C interpreter!
We felt right at home with Instant C. It has
the look and feel of C and the texture of a
BASIC interpreter, all at the same time.

Imagine instant gratification from this
language. Type "i+ +" and the value / is
printed, then incremented. Type " + +i"
and i is incremented, then displayed.
Learn C the easy way, with Instant C. We
debugged all of the DOS benchmark pro-
grams in Instant C before unleashing the
nasty old compilers on them.

The reason we did this is easy to under-
stand by looking at the compile times in
the benchmark results table— not a com-
pile (load) time over 8 sec. But not all can
be inferred from this table. The incredible
(200K!) interpreter includes an editor. If
you load a program with syntax errors,
the interpreter will instantly switch to the
edit mode and plunk the cursor on the
offending character. You won't be able to
exit this mode until you clean up all of
your syntax errors or cry uncle (quit).

Editing is very slick. Type "ED
< function name> " and only the function
appears, while the interpreter switches to
edit mode— modular text editing in the
context of an interpreter. Some very
admirable ideas arc written into this little
piece of code.

How about those error messages? "I'm
sorry, but I don't know the word 'num-
ber'.'' appeared when we had used "num-
ber" in one place and "NUMBER" in
another. Being a compiler means never
having to say you're sorry, but the inter-
preter seems to have that ability.

The bad news shows in the execution
time column in the benchmark results
table. Interpreters are slower than com-
pilers by a factor of two or three. Yet look
at the Matrix execution time. The inter-
preter approaches the speed of the com-
pilers in floating point intensive pro-
grams. This is somewhat intuitive, as a
floating point routine is the same the
whole world over, compiler or interpreter.

Deref was again the heartbreaker.
Above 16 indirection levels, the syntax is
too complex for Instant C's parser. At 16
indirection levels, the program ran, but
the memory-resident source was
destroyed. At 14 levels, all was well.

As to the source being clobbered— the
copy of Instant C we received was a pre-
release copy, version 0.9. This problem
will no doubt be fixed for the released
version. Let's hope so, this program is too
good to do without!

Unfortunately, you may have to do
without it. At $500, you get a great tool
for writing and testing functions and mod-
ules and nothing to compile them after
they are working. Rational Systems rec-
ommends that you purchase someone
else's compiler as a companion to Instant

COMPUTER tANGUAGE ■ FEBRUARY 1985

C, but wc think the combined purchase
price is abil steep.
On the other hand, if Instant C could

make you twice a.s productive, just think
what your company will save not having
to hire a second programmer. We sin-
cerely leel that Instant C can have a major
positive impact on programmer
productivity.

Mark William's MWC86

From the land of the minicomputer and
the super-microcomputer comes Mark
Williams Go's MWC86. Originally from
the UNIX environment (Coherent from
Mark Williams), MWC86 is a complete
development system for the MS-DOS
environment. It includes a compiler,
linker, assembler, librarian and symbolic
debugger.

MWC86 supports both small and large
models. In addition, floating point librar-
ies for non-8087 and 8087 use are
included. The compiler, linker, and one
library fit on a single disk.

The manual looks nice. It is typeset and
has fancy tab dividers. However, it is not
well indexed and doesn't seem to have any
error message explanations. The index
includes only function names and no top-
ics, so it is almost useless for the begin-
ner. The lack of error message expla-
nations is completely unforgivable.

The csd C-sourcc debugger is this pack-
age's claim to fame. Included is a vacuum-
formed template that fits over the IBM
PC's special function keys. The csd
debugger completely disassociates the
debug process from the assembly lan-
guage level, allowing the programmerto
deal only with C source code. The
debugger allows the program to run at full
speed, and program size does not grow
because the debugger is being used. On-
line help screens are also available, mak-
ing this the most comprehensive debugger
we have seen for MS-DOS C compilers.

During compilation of the benchmark
programs, wc encountered several errors
and warnings. As mentioned previously,
these were not explained in the manual
and we were left to our own resources.
Ultimately, we were able to compile all
four of the benchmarks, but Fib did not
run. In fact it hung the machine, making
us suspect MWC86*s capacity for
recursion.

Microsoft C (Lattice version)

It is somewhat common knowledge that
Microsoft C is actually Lattice C
(reviewed later). This isn't a secret as the
Microsoft manual lists Lattice as joint
copyright holder. Since the two compilers
sell for the same price ($500), it is very
interesting to compare and contrast the
two packages.

Microsoft C was supplied as version
2.03. This compiler supports the small,
program, data and large models. The
Microsoft C compiler does not support the

8087. but data formats have been made to
conform to the 8087 definitions.

No help is supplied for the creation of
the working disk. This is apparently left
as an exercise for the user. Microsoft sup-
plied all files on single-sided disks so
there was quite a bit of disk swapping
needed to create the work disks. No batch
riles were supplied to aid in compilation
or linking, but sufficient information was
in the manual for an experienced pro-
grammer to create these files. A copy of

K&R was included in the documentation.

The Microsoft C compiler is a full C
implementation. It was able to compile
and run the benchmarks with two mod-
ifications. The void typing in Fib had to be
eliminated, and Matrix required the addi-
tion of Mnclude <stdio.h>. In addition,
a fluke collision occurred between the
array c/llj and a procedure called c in the
standard program entry module. This

C Productivity Series—

The Professional's Edge

Blaise Computing has a
range of programming aids for
the most popular C compilers in
the IBM environment that no
serious system developer should
be without. These packages help
you lo easily access advanced
capabilities of the hardware and
operating system, and to finish
your projects with a substantial
savingof time and eFfort. With
software development costs and
pressures as great as (hey arc.
can you afford not to take advan-
tage of the finest tools available?

* C TOOLS" puts advanced
string handling functions at
your disposal and provides a
high-level interface to all
BIOS functions from your C
program. Complete screen
handling, graphics primi-
tives, and a substantial group
of useful, general-purpose
functions arc also featured. $125

♦ C TOOLS 2 + " lets your pro-
gram perform all the ad-
vanced DOS 2.0 services.
Program chaining, software
interrupt handling, and dy-
namic memory allocation are
all done "right." Buffer and
file handling functions are
provided, as well as a gen-
eral DOS gate. $100

C VIEW MANAGER"

our display screen manage-
ment system that makes
screen development and doc-
umentation much faster. It
comes with a complete
library of C functions which
use the screens you have de-
veloped to recall and display
information, capture and
validate field data entry, and
provide context-relevant
help files. S275
♦ ASYNCH MANAGER "is a
I ibrary of inte rrupt -driven
routines providing a general
interface to both COM ports
for your asy nchronous com-
munications applications.
Introductory price of 5175
includes all source.
AH of these products may be
used by developers with no
royally payments to Blaise Com-
puting. Source code either
comes with the package, oris
available. We support Lattice,
Computer Innovations, and
Microsoft C compilers. To expe-
dite your order or to obtain fur-
ther information, call or write us
directly.

Blaise Computing's Programmer
Productivity serin is also available in
versions far the Past nf fanguage.

BLAISE COMPUTING INC.

20A4 Blake Street Berkeley. C A 94704
(415) 540-5441

CIRCLE 11 ON READER SERVICE CARD

"This is a beautifully
documented, incredibly
comprehensive set of
C Function Libraries."

— Dr,>D6bb's Journal

ESJSSE

COMPLETE
SOURCES

[ PACK I: Building Blocks I

250 Functions: DOS,
Printer, Video, Asynch

I PACK 2: Database

100 Functions: B-Trees,
Variable Records

PACK 3: Communications

135 Functions: Smart-
modem™, Xon/Xoff,
Modem-7, X-Modem

PACK 4: Building Blocks II

100 Functions: Dates,
Text Windows,
Pull-down Menus
Data Compression

PACK 5: Mathematics I
35 Functions: Log, Trig,
Square Root

'-] PACK 6: Utilities I

Archive, Diff, Replace, Scan,
Wipe (Executable Files only)

Lattice™, Microsoft™, DeSmet™,
CI-86™ Compilers on IBM PC/XT/AT™
Small and Large Memory Models.
Credit cards accepted
(S7.00 handling/Mass, add 5%)

SOfTWfW

traizons
inc.

165 Bedford Street

Burlington, Mass. 01803

(617) 273-4711

NOVUM ORGANUM

CIRCLE 25 ON READER SERVICE CARD

84 COMPUTER LANGUAGE ■ FE8RUARY1985

resulted in a doubly-defined symbol error.
Changing the array name to f/r quickly
solved this problem.

Lattice C

It is apparent that Lattice C version 2 . 13 is
a later version of the compiler sold as
Microsoft C, version 2.03. Whole sec-
tions of the manual are duplicated in both
versions. Lattice C also supports the four
memory models: small, program, data,
and large. Several features have been
added to the Lattice version since Micro-
soft obtained the compiler.

The first addition is 8087 support. Lat-
tice C will automatically sense the pres-
ence of the 8087 coprocessor and use the
appropriate routines. Another addition is
in expansion of the library to include
more extensive mathematical functions. A
fork/waif capability has been added which
allows a program to create a child process
and wait for it to complete.

One addition we particularly appre-
ciated was an executive program called
LC. which manages the compilation of
programs. This replaced the batch file we
created for Microsoft C. Unfortunately,
no corresponding program is supplied for
linking, so we used the file developed for
Microsoft C, with suitable modifications.

We found the Lattice C manual to be
quite complex. It is haphazardly assem-
bled in three sections. The first section is
a supplement containing the additions in
version 2. 10 of the compiler. The large
middle section is the functional descrip-
tion of Lattice C, revision 2. The lust sec-
tion is a set of technical bulletins updating
the first section supplement with the addi-
tions made in versions 2.10, 2.11, and
2.13. Finding information in this con-
fusing array of pages is not very easy.
Unlike the Microsoft compiler, Lattice
does not include a copy of K&R with its
manual.

Ignoring the manual. Lattice C is the
most comprehensive compiler we have
reviewed for MS-DOS. Its memory
models, 8087 and library support provide
a very full-featured compiler. If Lattice
only supplied a debugger, it would be a
most complete development system.

Lattice C was able to compile all of the
benchmark programs, but Fib required
the deletion of the void typing, and Matrix
required the addition of ^include
< stdio. h > .

While linking Matrix we encountered
one problem — we got several error mes-
sages for missing functions. The manual
was not much help at first. Eventually we
discovered that Lattice had decided to
split the library, giving floating point
functions a separate library. Changing the
link command cleared this problem up.
The scattered documentation on the ver-
sion 2. 13 revisions were at the root of this
problem, as the other two manual sections
were quite misleading.

Microsoft C 3.0

This compiler is not at all like the C com-
pilers previously offered by Microsoft.
Before release 3.0, Microsoft sold
repackaged versions of the Lattice C com-
piler. With the release 3.0 package,
Microsoft introduces a homegrown com-
piler with very powerful features that take
very good advantage of MS-DOS. Of
course, this is to be expected since
MS-DOS is also a Microsoft product.

Microsoft has tried to reproduce a
UNIX-like environment with the release
3.0 C compiler. This can be seen in the
compiler environment and the library
functions supplied. Nothing has been
retained from the Lattice package. Micro-
soft does acknowledge its previous com-
pilers, however, by supplying a con-
version aid, a file called
V2TOV3.H.

The release 3.0 package is very large. It
has a four-phase compiler, a special
linker, several libraries, a librarian, and a
large collection of include files. The man-
ual is very large and comprehensive. It is
composed of three sections: a user's
guide, run-time library reference, and
language reference.

The manual recommends a two-disk
partition for floppy disk systems, placing
the compiler and include files on one disk,
with one set of libraries and the linker on
the second. Unfortunately, we couldn't
squeeze all of the include files on one disk
with the compiler. They must have gotten
somewhat larger since the manual was
written. The end result is that we highly
recommend a hard disk for use with this
compiler.

Microsoft has abandoned the Lattice
memory models in favor of the more pop-
ular Intel-style models. Release 3.0 sup-
ports small, medium, and large memory
models. In addition, Microsoft C release
3.0 supports the keywords near and far,
which can override the selected memory
model for data items. Memory model
selection is made by compiler command
line options.

We received a prerelease version of this
compiler from Microsoft. It was able to
compile the first three benchmarks with
no problems, but the Matrix benchmark
produced an internal compiler error. This
bug has been reported to Microsoft, and
we hope it will be fixed by the time the
compiler is offered for sale. The release
3 .0 compiler generated fairly compact
code that ran quickly for the Sieve and
Deref tests but suprisingly slow in the Fib
test. In all, we found the new Microsoft
compiler to be quite impressive and
expect it to be a contender in the crowded
MS-DOS C compiler market.

"THE SINGLE BEST DEBUGGER
FOR CP/M-80. A TRULY
AMAZING PRODUCT/'

LEOH ZOLMAN
AUTHOR OF BDS C

I ) Complete Upward Compatibil ty with DDT

D Simultaneous instruction, register, stack & memory displays

□ Software in-circuit-emulator provides write protected memory,
execute only code and stack protection.

□ Fifteen single keystroke commands

□ Thirty day money back guarantee

[~i On-line help & 50 page user manual

ONLY $195.
IBM PC VERSION AVAILABLE SOON!

-SOFTADVAIMCeS -

P.O. BOX 49473 AUSTIN, TEXAS 7876S (512) 478-4763

VISA

•We

\-m ■ : ■ -■ -m m n ■ ■ •

'* '" &$■* f^'"' ■■-*■■?

JB ■ ■ K H H- H'*

W : * M

'■■ -■ ■ ■■

CIRCLE 77 ON READER SERVICE CARD

WALTZ LISP™

The one and only adult Lisp system for CP/M users.

Waltz Lisp is a very powerful and complete implementa-
tion of the Lisp programming language. It includes
features previously available only in large Lisp systems, tn
fact. Waltz rs substantially compatible with Franz (the Lisp
running under Unix), and is similar to MacUsp. Waltz is
perfect for Artificial Intelligence programming. It is also
most suitable for general applications.

:h faster ihnn other microcomputer Lisps. • long integers (up to al 1 digits). Selectable radix • True dynamic
arocter strings. Full string operations including fast matching/extraction. ■> Flexibly implemented random file access.

■ Binary files. • Standard CP/M devfces. * Access to disk directories- * Functions a\ type lambda (expr), nlombda
{fexpr}, leicpr. macro. • Splicing and non-splicing character macros. • User control over all aspects, of the interpreter.

■ Built-in preitypnnling and formatting facilities. • Complete set of error handling and debugging functions including
user programmable processing of undefined function references. * Virtual function definitions. * Optional automatic
loading of irtif<ol»zalion fiSe. » Powerful CP/M command 'me oarsing. • Fast sorting/merging using user defined
comparison predicates. • Full suite of mopping functions, iterators, etc. * Assembly language interface. * Over ?50
functions in total. • The be*l documentation ever produced for o micro Lisp (300 -t full size pages, hundreds of
illustrative examples).

Version 4.4

Waltz Lisp requires CP/M-86 or CP/M 2.2, Z80 ond 48K RAM (more recommended). All
common 5" and 8" disk formats available.

r>0> («i

I'ro^^ode

-INTERNATIONAL—

*169

15930 SW Colony Pi.
Portland, OR 97224

Unix' Bell LabaralDnei.
CP/M- Digital Research Corp,

[Now includes Tiny Prolog
written in Woltz Lisp )

'Manual only: $30 (refundable with order). All
foreign orders: add $5 for surface mail, $20 for
airmail. COD add $3. Apple CP/M and hard sector
formats add $15.

Call free 1 "800-LI P-4000 Dept #1
In Oregon and outside USA call 1-503-684-3000

CIRCLE 60 ON READER SERVICE CARD

NEW FEATURES

(Free update for our early customers!)
» Edit & Load multiple memory

resident files.
» Complete 8087 assembler

mnemonics.
» High level 8087 support.

Full range transcendentals

(tan, sin. cos. arctan.

logs and exponentials)

Data type conversion and

I/O formatting.

• High level interrupt support.
Execute Forth words from with-
in machine code primitives.

• 80186 Assembler extensions for
Tandy 2000, etc.

■ Video/Graphics interface for
Data General Desktop Model 10

FORTH

• Fully Optimized & Tested for:
IBM-PC IBM-XT IBM-JR
COMPAQ EAGLE-PC-2
TANDY 2000 CORONA
LEADING EDGE

(Identical version runs on almost all
MSDOS compatibles!)

• Graphics & Text
(including windowed scrolling)

• Music - foreground and
background

includes multi-tasking example

• Includes Forth-79 and Forth-83

• File and/or Screen interfaces

• Segment Management Support

• Full megabyte - programs or
data

• Complete Assembler
(interactive, easy to use & learn)

• Compare

BYTE Sieve Benchmark [an 83

HS/FORTH 47 sec BASIC 2000 sec
w/AUTO-OPT 9 sec Assembler 5 sec
other Forths (mostly 64k) 70-1 40 sec
FASTEST FORTH SYSTEM

AVAILABLE.
TWICE AS FAST AS OTHER
FULL MEGABYTE FORTHS!

(TEN TIMES FASTER WHEN USING AUTO-OPT!)

HS/FORTH, complete system only: $250.

2S visa Mastercard £$
Add $10. shipping and handling

HARVARD
SOFTWORKS

PO BOX 2579

SPRINGFIELD, OH 45501
(513) 390-2087

CIRCLE 47 ON READER SERVICE CARD

THE j; JOURNAL

The C Journal will help YOU use c on your machine
-IBM PC", CP/NT, Macintosh '", or UNlX'"-based -
micro, mini, or mainframe.

it's the ONE publication for programmers, software
managers, and other computer professionals who
need to keep aware of developments in the indus-
try's fastest-growing language.

• regular columns for novice through advanced C
programmers

• software product and book reviews

• tips on working with major compilers and
operating systems

• news from the ANSI standards committee and the
industry

For FREE sample issue and discount subscription
information, write, call, or circle our reader service
number. The C Journal is a quarterly publication,
and costs $28/year (add $9 for overseas airmail).

Subscriptions/Advertising:

Christina Gardner

(201) 989-0570

Editorial:
Rex Jaesche

(703) 860-0091

another independent publication from ,— ^— |
E =-S=^= | Info Pro Systems ^°^

P.O. Box 849 \WZIM,

Denville, NJ 07834 *

CIRCLE 5 ON READER SERVICE CARD

OVERCOME
FORTRAN and PASCAL

LIMITATIONS WITH

LIMIT

Visa/MC

Alibraryof 58 Assembler routines transforms MS FORTRAN and PASCAL
plus other 3086/87/88 FORTRANs into the flexible, responsive, complete
language needed for the microcomputer environment. Featuring:

EXTENSIVE GRAPHICS (Get, Put, Paint, Color, Dot, Line, Box, Circle,
Ellipse, Large Characters)
(Windows, Cursor, Read/Write Screen)

(Match, Compare, Concatenate/Extract, Pack,
Justify, Zero Fill}

(Read Key Du ring Execution, String Read With
Edit)

(Exist?, Rename, Delete)

(Set Com Line, Send/Receive, Line/Modem
Status)

(Peek, Poke, Determine Time/Date, Random
Numbers, Beep, Clear Screen, OR/AND/
XOR/NOT/NEG of Byte/Word, Printer Status)
For routines to meet language specific problems
such as Chairing or Command Line Read call:

M | E | F Environmental Inc. P.O. Box 26537
Austin, Texas 78755 (512) 251-5543

FULL SCREEN CONTROL
STRING MANIPULATION

KEYBOARD CONTROL

FILE MANAGEMENT
COMMUNICATIONS

OTHER FEATURES

TESDCS TCCLS

HO DEI S II, 12, 16

TRSiC C COMPILER

Full K&R with source to the

function library. UNIX

compatible $85.00

ZSPF EDITOR

SPF, the choice of most

mainframe programmers, is

now available for Z80 machines.

And it's panel driven so you

can customize it! $75.00

MODEL IOC C CCMDIEED

Now you can write efficient programs for your TRS-80
model 100 with ease. Or, learn the essentials of C
programming while traveling!

C/100 - THE "PORTABLE" C COMPILER

Cassette version $49.00

Disk/Video interface version. . $59.00

Model II version (run on mod ll.then download

object code !o model lOO) $79.00

Model III version (as above, loi Mod III) $79.00

Write or call for information on other TRS-80 software.

TBSOOS and IRS 60 die Eiademarks o' T.m.!, CO'P

business utility snftuiare

IDS minna ste 4E3 san franciscn ca 94IDS

(415)397-2000

CIRCLE 55 ON READER SERVICE CARD

86 COMPUTER LANGUAGE ■ FEBRUARY1965

CIRCLE 17 ON READER SERVICE CARD

OPT-TECH SORT

SORT/MERGE program for IBM-PC & XT

Now also sods dBASE II files!

Written in assembly language for high performance
Example: 4,000 records of 128 bytes sorted to give

key & pointer file in 30 seconds. COMPARE!
Sort ascending or descending on up to nine fields
Ten input files may be sorted or merged at one time
Handles variable and fixed length records
Supports all common data types
Filesize limited only by your disk space
Dynamically allocates memory and work files
Output file can be full records, keys or pointers
Can be run from keyboard or as a batch command
Can be called as a subroutine to many languages
Easy to use, includes on-line help feature
Full documentation — sized like your PC manuals
S99 —VISA. M/C. Check, Money Order, COD. or PO
Quantity discounts and OEM licensing available

To order or to receive additional information
write or call:

OPT-TECH DATA PROCESSING

P.O. Box 2167 Humble. Texas 77347

(713) 454-7428

Requires DOS. 64K and One Disk Drive

mill i i — !^— saw 1

CIRCLE 66 ON READER SERVICE CARD

COMPILERS FOR CP/M

Recent years have witnessed a growing
support forC running under CP/M, start-
ing with Small-C and progressing to
ECO-C's highly professional imple-
mentation. This support has seeded a
large and continually growing body of
software written for the CP/M
environment.

J.E. Hendrix's Small-C
The original version of Small-C was pub-
lished in Dr. Dobb 's Journal in May 1980
by Ron Cain. In its original form, Small-
C was a simple one-pass compiler that
generated assembly language text for the
8080 processor. It recognized only char-
acters, integers, and one-dimensional
arrays of each. There were no Boolean
operators so bitwise logical operators
were used. The only loop control was the
while statement.

The compiler was updated and
extended by Jim E. Hendrix and pub-
lished again in Dr. Dobb 's Journal in
December 1982. The new compiler,
dubbed version 2.0 Small-C, has been
extended by the addition of code opti-
mization, data initialization, conditional
compilation, the EXTERN storage class,
the/or, do/while , switch , and goto state-
ments, assignment operators. Boolean
operators, the one's complement operator,
and assorted other features.

Since his article appeared, Jim Hendrix
has published The Small-C Handbook ,
which does an excellent job of documen-
ting the essential features of the compiler.
The book, in combination with a disk of
the compiler sources, would make an
excellent semester course on the structure
of seed compilers.

The Small-C compiler at $25 , with the
book S 14.95, brings this tool in at the low-
est price of any of the compilers we have
reviewed for CP/M. The combined price
tag of $39.95 is a good investment, as the
compiler makes an excellent tool for
doing small utility development in the
CP/M environment. Be aware that it
implements only a subset of the full
language.

Disk one contains the compiler
CC.COM, several submit files, a read. me
document, and order forms for the use of
people who wish to get their own copy of
Small-C. The read. me document deals
with a problem seen with using certain
versions of M80. The only instantly
apparent source file is called AR.C. We
displayed the source file and discovered
that it is a program archiver and the
assorted .ARC files on the disk are
archived sources for the system. A submit
file connects the compilation, assembly,
and link portions of building a program. It
seems as if almost everything has been
thought of. The documentation consists of
a copy of The Small-C Handbook .

The largest plus we see to Small-C is
that in combination with The Small-C

Handbook you have a ready tool to learn
about what a compiler is and how it
works. The compiler does not try to deal
with the entire language so it is somewhat
simpler and easier to read. Modifying the
compiler to extend the language is quick
and easy.

The Small-C compiler is in the public
domain and virtually free. What you are
paying for when you purchase the pack-
age is the right to be in on any updates and
fixes that come along, which is explained
in the file REGISTER on the disk. It is
your least expensive step into the C lan-
guage . If you try it out and find it's not to
your taste, you haven't made a heavy
investment. If you find that you like C,
then in the future you can move up to a
bigger implementation.

But be warned , do not expect to do
large-scale development in Small-C.
More comprehensive versions of the lan-
guage are out there. Many of the more
powerful extensions of C would make
some rather long and tedious Small-C
programs quite a bit shorter and more
concise.

Small C could run only one of the
benchmarks. Lack of floating point sup-
port and one-dimensional arrays pre-
vented the Matrix program from com-
piling, lack of unsigned integers kept Fib
from compiling, and lack of structures
stopped Deref.

Also be warned that if you do not
already have an M80 compatible relo-
cating assembler, and a linker, you are
going to have to get one in order to use
Small-C. Jim Hendrix does not supply
one with the package. Output of the com-
piler is assembly language source.

Small-C is not unique in offering the
entire compiler source as part of the com-
piler package. It was just the first. If you
need to fill a special niche in your tool
chest, then Small-C may fit the bill .

Software Toolworks' C/80

Software Toolworks' C/80 is a late entry
into the CP/M group of compilers. Fitting
between J.E. Hendrix's Small-C and The
Code Works' Q/C, it is the least expensive
of the C implementations offering a float-
ing point package. The compiler comes
with a configuration program and an
assembler. The total package with Math-
Pack and the compiler costs about $90.
For the price it seems to be a reasonable
implementation of C. However, it is not as
close to K&R C as the advertising would
like you to believe.

This compiler does sport some rather
impressive benchmarks, taking just 26 sec
to run the Sieve test. It then promptly
turned around and exploded on Deref. We
don't know what happened. It got lost and
filled the screen with garbage when

Pascal and C

Programmers

Your programs can
now compile the

FirsTime-

FirsTime is an intelligent editor that
knows the rules of the language being
programmed. It checks your statements
as you enter them, and if it spots a
mistake, it identifies it. FirsTime then
positions the cursor over the error so
you can correct it easily. FirsTime will
identify oil syntax errors, undefined
variables, and even statements with
mismatched variable types. In fact, any
program developed with the FirsTime
editor will compile on the first try.

Unprecedented

FirsTime has many unique features
found in no other editor. These powerful
capabilities include a zoom command
that allows you to examine the
structure of your program, automatic-
program formatting, and block
transforms.

If you wish, you can work even faster
by automatically generating program
structures with a single key-stroke. This
feature is especially useful to those
learning a new language, or to those
who often switch between different
languages.

Other Features: Full screen editing,
horizontal scrolling, function key menus,
help screens, inserts, deletes, appends,
searches, and global replacing.

Programmers enjoy using FirsTime. It
allows them to concentrate on program
logic without having to worry' about
coding details. Debugging is reduced
dramatically, and deadlines are more
easily met.

FirsTime for PASCAL $245

FirsTime for C 5295

Microsoft PASCAL Compiler $245

Microsoft C Compiler $395

Demonstration disk S25

Get an extra S100 offthe compiler when

it is purchased with FirsTime.

(N.J. residents please add O'.'fi sales tax.)

Spruce

Technology Corporation

I 10 Whispering Pines Drive

Lincroft, N..I. 07738

(201) 741-8188 or (201) 663-0063

Dealer enquiries welcome. Custom versions
for computer manufacturers and language
developers are available.

FirslTima is a iradi'mark r>r Sprues Technoiiigy

Oirpnralion. r T^^H i^i^i^H

VISA'

CIRCLE 33 ON READER SERVICE CARD

Thunder Software

• The THUNDER C Compiler- Operates und^r the APPLE Pascal ! I ap*tfsflngs^em.Createf*«tm[ tvnSl 12 programs ronjn
as Stand alone programs or as subroutine To PdwaiptogratUk ABaior*uKi.'t ol the CcJeiWd byKX; R Includes a 24 page users
gut eta. newsletters. Marm p^e^:>^:OkiasEH:"ra^ , : rn APP1JE If J t" 1 "- . «? -'V" Source code for I hraries Is Included Oniy $49*95

• ASSYST: The .WraWei Sytttem - A comptete6502 editor/ assembler and IisTtir for APPLE DOS3 3 Menu driven excellent
error trapping. 24 p uwrsgudt?. demo programs *<iuwe isode.ifls all programs! Gfed! lor be jiniv.'^ Only $23.50

• THUNDER XREF- A cross reference utility for APPLE Pascal 11 XRKI'iii^iLr.ttescro^sr^erertct ^l'irr<i. !i pTocedurt 3 Snimv
code and dncumerntiillnn provided Only Si 9.95

Thunder Software POB 31501 Houston Tx 77231 713-728-5501
Include S3.00 shipping. COD. VISA and MASTERCARD accepted

CIRCLE 65 ON READER SERVICE CARD

♦ Compilers:

CLC86 S349

Lattice S395

Mark Williams . . . $475

♦ General Libraries Si 75

(Lattice, Microsoft, Mark

Williams, CI C86)

♦ DeSmet C S150

♦ Comm Library ...... S 1 60

GREENLEAF SOFTWARE, INC.

2101 HICKORY DRIVE ♦ CARROLLTON, TX 75006 ♦ (214)446-8641

GREENLEAF

SOFTWARE r

CIRCLE 44 ON READER SERVICE CARD

MEMO: Qpxotfuvnmzu ^^

QUIT >^

WORKING

SO HARD,

These people have quit working so hard: IBM, Honeywell, Control Data,
GE, Lotus, Hospitals, Universities & Government Aerospace.

THE GREENLEAF FUNCTIONS™

THE library of C FUNCTIONS that probably has just what you need . . . TODAY!

. . . already has what you're working to re-invent

. . . already has over 200 functions for the IBM PC, XT, AT, and compatibles

, , . already complete . . . already tested ... on the shelf

. . . already has demo programs and source code

. . . already compatible wirh all popular compilers

. . . already supports all memory models. DOS 1.1, 2.0, 2.1

. . . already optimized (parts in assembler) for speed and density

. . . already in use hy thousands of customers worldwide

. . . already available from stock (your dealer probably has it)

. . . It's called the GREENLEAF FUNCTIONS.

Sorry you didn't know this sooner? Just order a copy and rhen take a break —
we did rhe hard work. Already.

THE GREENLEAF FUNCTIONS GENERAL LIBRARY: Over 200 functions in C
and assembler. Strength in DOS, video, string, printer, async, and system interface. All DOS 1
and 2 functions are in assembler for speed. All video capabilities of PC supported.
All printer functions. 65 string functions. Extensive time and date. Directory searches.
Polled mode async. (If you want interrupt driven, ask us about the Greenleaf Comm
Library.) Function key support. Diagnostics. Rainbow Color Text series. Much, much more.
The Greenleaf Functions. Simply the finest C library (and the most extensive).
All ready tor you. From Greenleaf Software.

. . . Specify compiler when ordering. Add $7.00 each for UPS second-day air. MasterCard,
VISA, check, or P.O.

attempting the pointer dereferencing. It is
hoped this problem will be fixed.

The source code for the lihrarics is
included in the package with the source
for the floating point package. Software
Toolworks has included at least some
rudimentary verification tools with the
package. It's a start at least.

We have a problem with the manual,
however. The 60 pages includes a 10-page
blurb on the MathPack. It covers the
topic, but is reminiscent of the original
CP/M user's guides that told you every-
thing you needed to know about CP/M, if
only you knew where to look. K.&R C and
C/80 arc going to have several points of
difference, so if your goal is portability,
this is not the too] for you.

Although this compiler docs not live up
to its advertised claim of being a full C
implementation, it docs have a place if for
no other reason than to create blazingly
fast utilities. Writing a large program
(over 10.000 lines) with this compiler
could be a trick. The main effort at least
initially would be to verify which con-
structs worked and which were going to
go out to lunch.

The Code Works' Q/C

The second compiler wc examined for
CP/M was the Q/C compiler from The
Code Works. This is the only other com-
piler wc know of that sells the compiler
package with source. (At least in the
affordable category.)

This compiler is a very good step up
from Small-C for the hobbyist. In particu-
lar, it has a respectably rich subset of the
C language. It differs in that it does not
support long .float . or double declara-
tions, which again means that Matrix can-
not be executed for this compiler. Lack of
support for extended-precision numbers
does not greatly reduce the value of this
compiler to the experimenter. The com-
piler sources are included.

Q/C. like the other less expensive com-
pilers, docs not implement parameterized
^defines . The Q/C compiler also does not
support the initialization of local, auto , or
register variables. Because Q/C is imple-
mented as a single-pass recursive descent
compiler, local variable declarations are
only supported immediately after a func-
tion header; local variables within blocks
and bit fields are not supported.

The preceding list is not particularly
extensive for a compiler in the SI 00
range, particularly one distributed in
source form. Educators needing a good
tool with its own programming course
should look long and hard at this product.

Wc particularly like tvvo features avail-
able with this compiler. The first is that
the compiler is available to create code for
one of two processors: 8080 or Z80. The
Z80 version generates only M80 (or
CWA— the optional Code Works assem-
bler) mnemonics because of optimizations
for the Z80 processor. The 8080 version

COMPUTER LANGUAGE ■ FEBRUARY19B5

oes not support Zilog mnemonics.
The second feature we liked is trace .
Vhen turned on, the compiler generates
ode that displays the name of each func-
ion when it is executed. This sounds
.andy for tracing down hard to find or
nvisible problems. The trace feature also
auses the compiler to search for trace
lags placed within comments and to gen-
rate the code necesary to output the text
vithin the trace flag text area.

The manual is excellent. It starts off
vith an introduction by the author. The
"irsl chapter shows a sample compilation
ising each of the recommended assem-
blers. It also covers the QRESET pro-
gram, which is used to change certain
parameters within the compiler (size of
:he symbol table, macro storage, default
Dptions, etc.). This is a reasonable way
of dealing with the limitations of a small
machine. The sources for QRESET are
part of the source library.

This manual, however, does have one
failing: it has no index. Most of the mate-
rial is well-organized, but certain things
are hard to find, such as information on
EXPAND.COM. This is not serious
because of the manual's good general
organization, just a nuisance.

The Q/C compiler, by Jim Colvin, has a
lot to recommend it. Sources are included
as part of the package. This means that
self-maintenance is possible for the hob-
byist, and the resource material made
available to the instructor makes it a must
acquisition. The price tag of close to $100
is not totally intimidating.

Some of the compiler's features (e.g.
trace) are unusually good ideas. We
would like to have this feature on some of
the VAX cross compilers we use at work.
In particular, the ability to embed trace
flags and debug text for output in com-
ments is one of those things that make you
hit yourself and say "Why didn't I think
of that?" Other features, such as the abil-
ity to compile code for multiple assem-
blers, indicate the author has run into the
new compiler syndrome: being able to do
the compilation but not having the assem-
bler to finish the job. Very frustrating.
Q/C is a small machine compiler. It
does what it is meant to do and it does it
well . The compiler is shipped set to com-
pile code modules with 200 symbols in its
symbol table. This can be changed. Still,
it indicates an inherent limitation of the
system. The limit is on the total number of
symbols per compilation rather than per
level of scope, which means you may need
to keep your modules small in order to use
this compiler effectively. We can't help
but wonder what the maximum limits on
these settings would be.

BD Software's BDSC

BD Software's BDS C has been used to
write software for the CP/M operating

system for about four years. In that time it
has become one of the most used and
respected C compilers within the CP/M
community. The compiler is written in
8080 assembly language and directly pro-
duces 8080 code. This is not always a
desirable feature, but in this case, because
it is a blindingly fast compiler, it is forgiv-
able. It is so fast that it can approach the
convenience of an interpreter for small
code fragments. BDS C accomplishes this
speed by loading the entire source file
from disk into memory and compiling in
place.

BDS C is the only CP/M C compiler we
have reviewed that has seriously exam-

ined the program verification and
debugging issue and has provided tools to
aid in this endeavor. This is accomplished
using two methods. BDS recognizes that
the CP/M environment has very effective
verification tools already available. These
tools are utilized by producing as an out-
put from the compiler a standard format
symbol file. This allows the the C pro-
grammer convenient access to assorted
symbolic debuggers supplied by other
vendors.

The compiler may also be instructed to
output 8080 restart (RST) instructions at

I/O A BORE? NOT ANY MORE!

ENHANCE

PROGRAM

INTERACTION

ELIMINATE

SCREEN ENTRY

ERRORS

COSTS

IF RE 1

~\_

REVISE SCREENS
IN SECONDS

M|E(F Environmental, Inc.

P.O. Box 26537

Austin, TX 78755 (512)251-5543

S450 : OEM Pricing and Licenses Available

Demonstration Diskette S 10

applicable to purchase

CIRCLE 15 ON READER SERVICE CARD

CODE SIFTER

THE EASY TO USE
SYMBOLIC EXECUTION PROFILER

No knowledge of assembly language is needed.

Easy to use - become productive in minutes.

Online HELP at each menu complements the user
manual.

I Define 32 ranges in USER/DOS/BIOS with a single
keystroke or enter them manually if you desire.

Use the looping feature to isolate the busy areas of
your program automatically.

Symbolic statistical output can be sent to any device
or file.

Adjustable sampling rate for increased accuracy.

Free demonstration diskette available, (requires deposit)

David Smith Software

Box 25 A R,D.#3

Oxford, N.Y. 13830

(607) 843-6209

Requires IBM-PC
128K DOS 2.X

$119

IBM is a trademark of International Business Machines Corp.

CIRCLE 36 ON READER SERVICE CARD

Eco-C Compiler

Release 3.0

We think Rel. 3.0 of the Eco-C Compiler is the
fastest full C available for the Z80 environment.
Consider the evidence:

Benchmarks*

(Seconds)

Benchmark

Eco-C

Aztec

Q/C

Seive

Fib

125

Deref

CNC

Matmult

115

N/A

'Ttmes courtesy ol Or David Ctark
CNC - Could Not Compile
N/A - Does not support floating point

We've also expanded the library (120 func-
tions}, the user's manual and compile-time
switches (including multiple non-fatal error
messages). The price is still $250.00 and
includes Microsoft's MACRO 80, As an option,
we will supply Eco-C with the SLR Systems
assembler - linker - librarian for S295.00 (up to
six times faster than MACRO 80}.
For additional information,
call or write: .

(317) 255-6476
6413 N. College Ave. • Indianapolis, Indiana 46220

the beginning of each C statement. This
allows CDB, BDS's symbolic C debugger,
to directly control execution of the target
program by setting breakpoints and allow-
ing symbolic display and modification of
variables. This is definitely a step toward
reducing the drudgery of program
verification.

The ability of BDS to quickly compile
and integrate the resulting code into a
very good symbolic debugger has gone a
long way toward making BDS C one of
the preferred tools in the CP/M program-
mer's toolchest.

The release wc reviewed (version
1 -50a) still suffers from a lack of real
number implementation. It does, how-
ever, support a floating point function
library, which is not integrated at all into
the compiler. Thus, float, double, and
long data types are not directly supported.
Functions similar to the intrinsic long and
float operations are available from the
library using byte arrays.

The C Users Group (formerly the BDS
C Users Group) is one of the more active
software-producing users groups in the
country. The. float and longint package as
well as the CDB symbolic debugger
resulted from efforts by members of this
group.

Library sources have been included to
allow customization by users for non- CP/
M environments. Along with these are a
large number of source files. The CDB
symbolic debugger is included in source
form. There is a library for directed I/O, a
utility for command line wild card expan-
sion, an assembler to assemble to BDS
C's own .CRL-format relocating libraries
and the float and longint libraries men-
tioned previously. Othello (a game!),
several file utilities, and finally Teledit, a
large telecommunications program, are
also included.

BDS C is one of the best values for an
all-around useful tool. For $150 you have
the fastest CP/M compiler on the market.
If you thought that program turnaround
was not important, then you have never
had to do serious debugging. Given the
difficulty of system development efforts,
we want all the help we can get in building
and verifying software. We wish more
compilers used the intelligence and acute
sense of reality that BDS C has placed in
their debugging tools. Many of the
mainframe-based cross development com-
pilers have some form of hook into the
compiled code, which allows a debugging
system to be integrated into the software.

Because of the in-memory compilation,
BDS C is best used for module develop-
ment. The size of the source code body
that can be compiled is limited to 48K or
so in a 64K CP/M environment. In certain
cases this may impose restrictions on the
size and number of functions available in
any one compilation.

CIRCLE 22 ON READER SERVICE CARD

90 COMPUTERLANGUAGEBFEBRUARY1985

The greatest failing of the BDS C com-
piler has been the .CRL libraries' non-
standard format. Some sort of conversion
capability that would convert between
.REL format and .CRL format libraries
would be an excellent addition to the sys-
tem. The new CASM assembler is a good
addition to the package. It just fails in
many ways to link with the large body of
excellent development tools available
under CP/M.

AlcorC

Alcor Systems' Alcor C is one of the two
CP/M implementations of C in this article
that could be termed a fitll imple-
mentation. It has the most comprehensive
manual of all of the implementations that
we have reviewed. It is divided into six
sections: beginners, editor, system,
tutorial, reference, and Advanced Devel-
opment Package (ADP). It contains about
470 pages of information about the use of
the compiler and editor. It is well laid out
with a table of contents and index for each
section.

The package is divided into three main
systems and the editor. The first system is
the compiler and the p-code interpreter.
The second system is the ADP. It contains
the p-code-to-native machine conversion,
the p-code optimizer, and the LinkLoadcr,
which allows both native machine code
and optimized p-code functions to be used
in the same program.

The LinkLoader allows you the ability
to profile the program and convert those
portions requiring high efficiency to
native machine code while still retaining
the size efficiencies of p-code for the
majority of the program. The ADP pack-
age is offered as an option to the base
compiler, but for the serious development
effort it is a must. The third section is
Alcor's XASM multi-assembler. The
assembler is not an essential portion of the
system, but the documentation on the
assembler contains a complete description
of Alcor's low-level p-code. This is essen-
tial fordoing large-scale development in
this type of environment.

If your development effort is concerned
with porting software, this compiler is
one of the two CP/M C compilers
reviewed that can do the job. Alcor C is
very close to being full K&R C. Most of
the differences from K&R are extensions
on the base language. Alcor C has only
two restrictions: the address of an array
element may not be used as an initializer
and external float and double variables
may not be initialized.

It becomes apparent very quickly that
p-code implementations have many
advantages. One of them is the com-
pletenessofthe implementation. The
Alcor C language implementation is one

PERFORMANCE
PORTABILITY

team
I SAN LPZ^GC3Za\(3l1

gj3£]ffi

UNBEATABLE

ctree

BY FAIRCOM

2606 Johnson Drive
Columbia MO 65203

The company that introduced micros to
B+ Trees in 1979 and created ACCESS
MANAGER 1 " for Digital Research, now
redefines the market for high performance,
B+ Tree based file handlers. With c-tree'"
you get:

• complete C source code written
to K&R standards of portability

• high level, multi-key ISAM routines
and low level B+ Tree functions

• routines that work with single-user
and network systems

• no royalties on application programs

$395 COMPLETE

Specify format:
5'/<" PC-DOS V/i" Mac
8" CP/M* 8" RT- 1 i

for VISA, MC or COD orders, call

1-314-445-6833

Access Manager and CP/M are trademarks of Digital
Research, Inc.

e-tree and the circular disc logo are trademarks
of FairCom

<B 1984 FairCom

CIRCLE 29 ON READER SERVICE CARD

:<§

TOTAL CONTI

FORTH: FOR Z-80@, 8086, 68000, and IBM®

Complies with the New 83-Standard

GRAPHICS* GAMES .COMMUNICATIONS •ROBOTICS

DATA ACQUISITION • PROCESS CONTROL

• FORTH programs are Instantly
portable across the four most popular
microprocessors.

• FORTH is interactive and conver-
sational, but 20 times faster than
BASIC.

• FORTH programs are highly struc-
tured, modular, easy to maintain.

• FORTH affords direct control over
all interrupts, memory locations, and
i/o ports.

• FORTH allows full access to DOS
files and functions.

• FORTH application programs can
be compiled into turnkey COM files
and distributed with no license fee.

• FORTH Cross Compilers are
available for ROM'ed or disk based ap-
plications on most microprocessors.

Trademarks IBM, Internalional Business Machines
Corp. CP/M. Digital Research Inc. POForih + and
POGEN, Laboratory Microsystems, Inc

FORTH Application Development Systems

include interpreter/compiler with virtual memory
management and multi-tasking, assembler, lull
screen editor, decompiler, utilities and 200 page
manual. Standard random access hies used lor
screen storage, extensions provided for access to
all operating system lunctions
Z-80 FORTH lor CP/M" 2.2 or MP/M II, $100.00.
8080 FORTH 'or CP M 2.2 or MP/M II. $100 00,
8086 FORTH for CP/M-86 or MS-DOS, $100.00.
PC/FORTH tor PC-DOS. CP/M-86. or CCPM,
$10000: 68000 FORTH for CP/M-68K. S250.00

FORTH + Systems are 32 bit implementations
that allow creation of programs as large as 1
megabyte. The entire memory address space ol
the 68000 or 8086/88 is supported directly.

PC FORTH + S250.00

8086 FORTH + lorC=/M-86 or MS-DOS S250-00
68000 FORTH + for CP/M-68K $400.00

Extension Packages available include' soft-
ware floating point cross compilers, INTEL

8087 support. AMD 951 1 support, advanced col-
or graphics, custom character sets, symbolic
debugger, telecommunications, cross reference
utility. B-tree file manager. Write lor brochure

Laboratory Microsystems Incorporated

Post Office Box 10430, Marina del fley, CA 90295
Phone credit card orders to (273J 306-74T2

CIRCLE 35 ON READER SERVICE CARD

which we would not hesitate to use in a
large programming effort. The profes-
sionalism presented is very apparent.

One of the elements that seems to be
missing, however, is effective support for
program verification. Since the compiler
produces p-code that is interpreted at one
level in the development, one would
believe that a very good symbolic
debugger could be produced. In particu-
lar, other languages are available from
Alcor running the same p-codes, which
would make the development of an inter-
active debugging tool to go with its lan-
guage packages quite attractive.

The p-code implementation also allows
for a large system development on a small
machine. This may seem to be con-
tradictory at first, but consider that the p-

code, instruction for instruction, is much
more dense than native machine code. So
that once the program size has made up
for the interpreter overhead, the p-code
begins to be much more compact than the
native code implementation.

However, for small programs and util-
ities, p-code loses big. Even with p-code
conversion to native code, the native code
compilers will win out. Interpreted
languages— whether text-based or p-
code-based— all have one failing: they
bring one more level of complexity and
potential for error into the picture.

The p-code interpreter within the Alcor
package is murderously slow. The Sieve
benchmark took 3.3 min to run. We abor-
ted Fib after 5 min. This seems to be
about 10 times slower than most of the
native code implementations. Even with a
3-4 times improvement in speed of gener-

A POWERFUL 68000 DEVELOPMENT
ENVIRONMENT FOR YOUR Z80 SYSTEM

C01668 ATTACHED RESOURCE PROCESSOR

• 68000 Assembler

• C Compiler

• Forth

• Fortran 77

Pascal

BASIC-PLUS
CBASIC
APL. 68000

6 MHZ 68000 CP/M-6SK 768K RAM

4 x 16081 MATH CO-PROCESSORS CPM80 RAM DISK

Develop exciting 68000 applications on your current Z80 based CPM system using
powerful mini-frame like 32 bit programming languages. And then, execute them at
speeds that will shame many S100K plus minicomputer systems.

The C01668 ATTACHED RESOURCE PROCESSOR offers a Z80 CPM system owner a
very low cost and logical approach to 68000 development. You have already spent a
small fortune on 8 bit diskette drives, terminals, printers, cards cages, power
supplies, software, etc. The C01668 will allow you to enjoy the vastly more powerful
68000 processing environment, while preserving that investment.

C01668 ATTACHED RESOURCE PROCESSOR SPECIAL FEATURES:

• 68000 running at 6 Mhz

• 256K to 76SK RAM (user partitioned between
CPU and RAM Disk usage)

* Up to four 16081 math co-processors

• Real time clock. 8 level interrupt controller
& proprietory I/O bus

* Available in tabletop cabinet

* Delivered w/ sources . logics, & monolithic
program development software

• Easily installed on ANY ZS0 CPM system

• CP/M-68K and DRPs new UNIX V7 compatible
C compiler {w/ floating point math) - standard
feature

• Can be used as 76SK CPM30 RAM Disk

• Opiional Memory parity

• No programming, or hardware design required
for installation

• Optional 12 month warrantee

• PRICES START AS LOW AS $899.00 for a C01668 with 256K RAM. CPM68K, C Compiler, Sources.
Ptints, 200 page User Manual. Z80 Interface, and 63000 System Development Software.

For further Information about this revolutionary product or our Intel 80S6 Co-Processor, please send Si
[no checks please] or call;

Hallock Systems Company. I nc .
262 East Main Street
Frankfort. Ntui York 13340
(315) 895-7426

RESELLER AND OEM

INQUIRIES INVITED.

CIRCLE 31 ON READER SERVICE CARD
92 COMPUTER tANGUAGEB FEBRUARY I985

ated native code, it is running 2-3 times
slower than the competition.

Ecosoft's ECO-C

Ecosoft proudly states that the only
incompatibilities that its Eco-C has with
the full C syntax are lack of bit fields and
the Mine macro directive. The Eco-C
package provides everything that is
required to compile and run programs. In
particular, it provides one of the best
assemblers we have stumbled across, the
SLR assembler by SLR Systems. The
inclusion of tools of this quality does not
dim the quality of Ecosoft's C imple-
mentation but complements it. creating a
very good program development tool.

Eco-C is a very clean implementation
of the C language. Portability of C pro-
grams from other systems is very good.
The SLR assembler selected as a compan-
ion tool to the compiler is a breath of fresh
air to the assembly language programmer.

The manual for Eco-C is 1 15 pages
long in an off-size format binder. For the
large-scale project, you should see if you
can order several manuals al extra cost. It
is very concise and professionally done,
concentrating on the essentials of oper-
ating the compiler, creating libraries, the
assembler interface, and Ecosoft's ver-
sion of the standard library interface. The
compiler docs not include any sources for
the STDIO but docs state that Ecosoft will
make them available for a simple copying
charge and release statement.

Included in the same binder is the man-
ual for the SLR assembler. Approxi-
mately the same length as the compiler
manual, it quickly illustrates the use of the
assembler, linker, and librarian included
in the package. Just as an example, the
assembler can be configured from the
command line to output cither a .COM
file, a .HEX file or one of two separate
.REL file formats. We have not even dis-
cussed how much faster it executes than
M80. Too bad we are not reviewing
assemblers!

At $250, the Eco-C package is out of
the range of most individual program-
mers. Even so, with the inclusion of the
SLR assembler for S45 extra, it is a good
investment. The compiler and assembler
are good for a large range of software
development, from the small utility devel-
opment effort to massive multi-
programmer efforts. This is the only
package we reviewed that we would be
willing to call a professional tool. The
compiler options, although not simple,
were well-documented. The benchmarks
were not the fastest but were in the ball-
park, making this a fine tool for all-
around use.

The compiler is such a nice tool that it
is too bad program verification interfaces
consist entirely of using the .SYM output
file al the assembly language level. There
should be some way of hooking a sym-
bolic debugger to the final program,

cither as a compiler option to output
restart instructions or hooks to allow for
post processing the assembler output (i.e.,
output line numbers and line boundary
markers as a compiler option).

OS-9C COMPILERS
Microware C

Microware offers one of the few C com-
pilers for the OS-9 operating system and
the 6809 processor. That's not too much
of a surprise because OS-9 is a Microware
operating system. In case you haven't run
into OS-9 before, it is a multi-user and
multitasking operating system for the
Motorola MC6809 8/16-bit micro-
processor. OS-9 is blessed with UNIX-
like tree directories and can support over
1MB of system memory with memory
management. It is therefore a very com-
fortable environment for the C language.

Though it is very close to a full C
implementation. Microware C differs
somewhat from K&R C. Bit fields arc not
supported. Constant expressions for ini-
tializers may include arithmetic operators
only if the operands in the expression are
of type int or char . The #if < constant
expression > is not supported although
ffifdef, ffijhdef. if else and ftendif are.
Macro definitions and strings must reside
on a single line of source code. Finally.
the S n escape sequence is output as a car-
riage return, which is not interpreted as an
"end of line" character in most systems
but is exactly what OS-9 requires.

On the positive side, Microware has
added some interesting enhancements. A
new storage class. Direct, has been added
to take advantage of the 6809's direct page
register. Though use of this feature will
make a program non-portable, it can be
used when program execution speed has
the highest priority.

In addition, Microware's C supports in-
line assembly code with the #asm and
Mendasm preprocessor commands. It is
possible to output a linefeed using the \ I
line feed escape sequence extension.

The Microware level II compiler com-
prises six separate programs. An exe-
cution manager, called CC2, is supplied
as one of these programs. CC2 invokes
the other five programs, simplifying the
programmer's interface. The compilation
sequence involves the preprocessor, com-
piler, optimizer, assembler, and linker.
Also included in the package is an exten-
sive library of header files that facilitate
operating system calls, a Microware
BASIC09 interface, and several other use-
ful functions.

Intro! C

Introl offers the other C compiler for the
OS-9 operating system reviewed in this
article. It is available for the Flex and
Uniflex operating systems as well. Introl
C is almost a full K&R implementation,

lacking bit fields, double float variables
and the Mine and ft/preprocessor
directives.

Introl's compiler has some enhance-
ments. Comment nesting is permitted,
allowing the programmer to comment out
a section of code by bracketing the code
segment. Separate name spaces for each
union and structure are supported so that
identical names can be used in different
struct and union declarations. Symbols
can be up to 90 characters in length.

For $425 Introl supplies the compiler, a
relocating assembler, linker, a library and
a librarian. This comprises a nearly com-
plete OS-9 language development system,
lacking only an editor. For the hardware
developer, a library of modules is avail-
able for generating stand-alone, ROMable
code. Introl's documentation is superb.

One amenity Introl C lacks, however, is

a module to invoke the various compiler
phases automatically. With preprocessor,
parse, code generation, optimizer, assem-
bler and link phases all required to gener-
ate an executable file, such a utility would
be most welcome. Introl's C compiled
more slowly than Microware's compiler
and generated larger executable files.
However, this is offset by the faster exe-
cution times of Introl C's compiled
programs.

CP/M 86 C COMPILERS

The last MS-DOS C compiler we
reviewed is from Digital Research, which
submitted both versions for review. It is
similar to the CP/M 86 version, so we dis-
cuss both in this section.

CP/M 86 has not done well in the IBM

Realia COBOL
Numbers speak
louder than words.

Compilation Speed (minutesiseconds)

Lines in
Program

Realia
COBOL

mbp
COBOL

Level II
COBOL

R-M
COBOL

Microsoft
COBOL

1,000

:51

8:33

3:42

5:05

5:11

5,000

3:30

48:07

16:58

45:26

"Could not successfully compile the program.

Execution Time Ratio

(Gibson Mix; calculated S-Profile)

Realia
COBOL

mbp
COBOL

Level II
COBOL

R-M
COBOL

Microsoft
COBOL

1.0

3.6

14.7

21.6

22.3

Sieve of Eratosthenes

0.818 seconds per iteration

All benchmark tests were performed on an
IBM PC-XT with 192KB of memory. IBM
PC-XT is a registered trademark of Inter-
national Business Machines Corporation;
mbp COBOL, of mbp Software and Sys-
tem Technology; Level II COBOL, of Micro
Focus; R-M COBOL, of Ryan-McFarland;
and Microsoft COBOL, of Microsoft.

+ IBM VS COBOL compatibility.

REALIA

$995

10 South Riverside Plaza
Chicago, Illinois 60606
(312) 346-0642

CIRCLE 76 ON READER SERVICE CARD

F77L, THE FORTRAN

THAT CHALLENGES

THE BIG SOFTWARE

COMPANIES.

When you have one software product to
sell, you had belief make it count. F77L, our
complete implementation of the ANSI
FORTRAN 77 Standard for the IBM PC, is for
programmers who buy a language system
based on features and performance.

By specializing in FORTRAN, we see
ourselves as different from our competition.
Rather than be jacks of all languages. LCS
prefers to be experts in FORTRAN. O'jr
competitors, on the other hand. lend to be
large and ferocious software firms that offer
computer users an assortment of products.
At LCS. we specialize in FORTRAN. We have
been successfully implementing FORTRAN
language systems for over 15 years. We may
not have the name recognition or the
advertising buctgetsof the big firms, but we
do have what matters most to you — a great
product.

Here are a few of the many reasons to buv
F77L:

• Full F ORTRAN 77 Language with popular extensions.

• Compile time: multiples faster than any
Goliaths'. ilOO's statements/min.)

• User oriented interface.

• Numerous, specific English level diagnostics
displayed during compilation,

• Command Line compiler options.

• Execution error tracebark: program unit line
number.

• Selective protection lor constants, bounds,
interfaces.

• Standard or Free Formal source files.

• Lattice C compatibility.

• Easy to follow manual includes appendices
on interfaces to Lattice C and Assembly
language.

• 30-day-money-back guarantee and ongoing
user support.

If you're tired of betting on the software
Coliaths and losing, call LCS. the FORTRAN
specialist.

$477 for complete package: one 5V4"
floppy and manual. Visa/MC Multiple copy
discounts. "Requires: 256K/8087.

To order or for more information:

Lahey Computer Systems, Inc.

904 Silver Spur Road, Suite 417
Rolling Hi lis Estates. CA 90274
213/541-1200
Serving the FOR TRAN community since 7969

IBM is a Ifddemjrk utf IBM corporation

Latlue C *s a trademark of Lattice, Inc.

PC segment of the 16-bit market but has
fared better with the other computers
using the 8086 processor. Since CP/M 86
is a Digital Research operating system, it
is of no surprise that the one CP/M 86 C
compiler tested is also from Digital
Research, though many of (he MS-DOS C
compilers are also available in CP/M 86
versions.

Digital Research C
We had a bit of trouble coaxing the CP/M
86 version of Digital Research C to run on
our IBM PC. This was probably not the
compiler's fault. We had an old copy of
CP/M 86 (version 1.0), and the compiler
simply returned to the operating system
before completing the compilation. Ver-
sion 1 . 1 of CP/M 86 cleared this up, and
wc were off and running.

Digital Research C supports the small
and big 8088 memory models. Memory
model selection as well as 8087 support
are specified at compile time by command
line options.

One very nice feature of the Digital
Research C compiler documentation is the
inclusion of a copy of K&R with the man-
ual. Digital Research made it quite clear
how to configure a working disk with
compiler, linker, and library. We had no
problem creating one.

Source code was another story. Our
benchmarks had all been written using a
text editor under MS-DOS, and we had no
editor for CP/M 86. Copies of CP/M 86
compatible editors are few and far
between in our area, and wc turned a bit
green when we considered using the
CP/M 86 editor ED to create the files.
Fortunately we have a very handy utility
program called Xenocopy Plus from Ver-
tex Systems, which promptly transferred
the four benchmark source files to a
CP/M 86 disk.

Sieve compiled and ran with no prob-
lem. Fib required that we add ttindude
<portab.h > so that the void typing
would be recognized. Deref and Matrix
also compiled and ran fine.

An interesting aspect of the compilation
is that the compiler places information in
the object file so that the linker knows
which libraries to use. No library parame-
ters are in the link syntax. The compiler
was not smart enough to place the object
file on the same disk as the source file. It
required we specify we wanted this by
using the -z compiler option. However,
the linker knew that we wanted the exe-
cutable file on the same disk as the object
file. Consistency once again is desirable
but lacking.

Digital Research also submitted for
review the MS-DOS version of its C com-
piler. Its benchmark performance is
included in both the MS-DOS and CP/M
86 benchmark tables for comparisons.
Two very interesting conclusions are that

Digital Research's compiler generates
very fast code, consistently showing one
of the best execution times in the
MS-DOS category.

The other point of note is that the
MS-DOS version of the compiler gener-
ates programs more quickly than the
CP/M 86 version. Though not shown in
the tables, we checked the times required
to compile and link separately. The
MS-DOS versions of both the compiler
and linker are faster than their CP/M 86
counterparts. We suspect that this is actu-
ally better performance on the part of
MS-DOS during floppy accesses.

TRS-80C COMPILERS

Three C compilers are available for the
TRS-80 Z80 computers: LC, Alcor C,
and Aztec C. Because other machine ver-
sions of the Alcor and Aztec C compilers
have already been covered, we will focus
here on MiSosys's LC compiler.

LC is based on Small-C and, like
Small-C, is not a full implementation of
the language. However, LC provides
more of C's features than does Small-C.
The switch and goto statements are
included, greatly simplifying command
parsers and similar routines. The omis-
sions (floating point capability, long
integers, unions , and structs) are signifi-
cant only if you try to use it as a number
cruncher.

LC first appeared on the scene in
December 1982 and has undergone only
two significant changes since then. The
first was a general bug-fix shortly after
the package was released, which brought
the version number to 1 . 1 ; the second
actually did not affect the compiler but
rather greatly extended the supplied
libraries. This, the current version, is 1 .2.
This review was done using version 1.1.

When you buy LC (for SI 50), you get
two disks (and a large 3-ring binder filled
with an excellent user manual). One disk
contains the LC compiler itself together
with the libraries, an /ASM skeleton, and
a /JCL file. The other contains EDAS, a
full-featured macro assembler (which can
also be purchased separately). The LC
strategy is to compile the C source code
into an assembly-language file, which is
then assembled by EDAS. EDAS pro-
duces executable code as its output; no
linking step is required.

On TRS-80 model 1 or 3 , LC programs
require the LDOS operating system. No
other DOS provides the capability for I/O
redirection. On model 4, LC programs
will run under all versions of TRSDOS 6.
Incidentally, by using the appropriate
/ASM skeleton and libraries, LC pro-
grams can be cross compiled between
models 1 , 3, and 4. Any model 1/3 LC
program will run on cither model inter-
changeably ; model 4 programs interface
to the hardware differently but can be
compiled by the model 3 compiler.

The assembly-language output from LC

CIRCLE 62 ON READER SERVICE CARD
94 COMPUTER LANGUAGE! FEBRUARY 1985

Scroll & Recall

Screen and Keyboard Enhancement

for the IBM - PC, XT and Compatibles

Allows you to conveniently scroll
back through data that has gone off
the top of your display screen.

Allows you to easily recall and edit
your previously entered DOS com-
mands and data lines.

Very easy to use, fully documented.
Compatible with ail versions of DOS,
monochrome & graphic displays.

$69 - Visa, M/C, Check, COD, POs
Phone orders accepted

Make Your Work Easier!

To Order or to Receive Additional

Information, Write or Call:
Opt-Tech Data Processing

P.O. Box 2167 • Humble, Texas 77347
(713) 454-7428
Dealer Inquiries Welcome
-*—

CIRCLE 67 ON READER SERVICE CARD

/ SUPE^ORTH 64 "" ^^

TOTAL CONTROL OVER YOUR COMMODORE-64'"
USING ONLY WORDS

MAKING PROGRAMMING FAST, FUN AND EASY!

MORE THAN JUST A LANGUAGE .. .

A complete, lully-integrated program development tystem.

Home Use, foil Gomel, Graphics, Data Acquisition, Business

Red Time Process Control. Communications, Robotics, Scientific. Artificial Intelligence

A Powerful Supenet ol MVPFORTH/ FORTH 79 * E»f. for the beginner or profeiiionol

e SPRITE-EDITOR

e Access oil C-64 peripherals including 4040
drive

e 20 to oOO X falter than Basic

• 1/4 * Ihe programming time
e Easy full control ol all sound , hi res.

gra phics , color, sprite, plotting line &
circle

• Controllable SPLIT-SCREEN Display

e Includes interactive interpreter & compiler

• Forth virtual memor y

• Full cursor Screen Editor

• Provision lor application program
diltribution without licensing

• FORTH equivalent Kernal Routines

• Conditional Macro Assembler

• Meeft all Forth 79 standards*

• Source screens provided

e Compatible with the book "Storting Forth'
by Lea Smote

• Access to oil I/O ports RSI32. IEEE,
including memory & mteinjptt

e ROMABLE cade generaror
» MUSICEDIfOB

SUPER FORTH 64' » --■■

powfivl than m«l ofihfr (prnpuirf btiffL.OB»' J
» SUFERFQRTH64 »

• 5 ngle disk drive backup utility

• Disk & Cassette bated. Disk included

• Full disk uioge — 680 Sectors

• Support* all ComTodore file typei and
Forth Virtual disk

• Access to 2Q< RAM underneath ROM
a reai

• Vectored kernel words
t TRACE facility

• DECOMPILER facility

• Full Siring Handling

• ASCII error meisagei

■ FLOATING POINT WtlH SIN/COS A SORT

• Conversational ui#f defined Command*

• Tutorial eiamples provided, In extensive
monuol

• INTERRUPT routines provide easy control
of hardware tirnen, alarms and devices

• USER Support

SUPER FORTH 64 «™,w «d.

.-.c;; - ■-■■ m-or» compocf than **ei atttrr-bly cede!

BASfC
FORTRAN
ASSEMBLER

A SUPERIOR PRODUCT

in ovary way! At a low

price of only

$96

Power ol Languages Constliucts Program Functionality

Call:
(415) 651-3160

PARSEC RESEARCH

Drawer 1776, Fremont, CA 94533 ■ —;

- unit ■( a* >; h i,,w„ <*n ChmMpim *j t vc-w H» «l Qtmmitm o> c o.d

Tdkt Ih.i ad to (our k<a

■Je-li *' <t' I D0h»>l sjook-

iToit PKifjrii Wifin olw

MicroMotion

MasterFORTH

It's here — the next generation

of MicroMotion Forth.

CIRCLE 49 ON READER SERVICE CARD

• Available tor Ihe APPLE IPs, IBM PC, Macintosh &
CP/M 2.x

• Meets all provisions, extensions and experimental
propose Is of the FORTH-63 1 nternational Standa rd ,

• Uses the host operating system .

• Built-in micro-assembler with numeric local labels.

• A full screen editor is provided which includes 16 x
64 format, can push & pop more than one line,
user definable controls, upper/lower case key-
board entry. ACOPY utility moves screens withln&
between lines, line stack, redefinable control
keys, and search & replace commands.

■ Includes all file primitives described in Kernigan
and Plauqer's Software Tools.

• The inputandoutputstreamsarefullyredirectable.

• Theeditor.assemblerandscreencopyutilitiesare
provided as relocatable object modules. They
are brought into the dictionary on demand and
may be released with a single command.

• Many key nucleus commands are vectored. Error
handling, number parsing, keyboard translation
and so on can be redefined as needed by user
programs. They are automatically returned to
their previous definitions when the program is
forgotten.

• The string-handling package is the finest and
most complete available.

• A listing of the nucleus is provided as part of the
documentation.

• The language implementation exactly matches
the one described in MASTERING FORTH , by Ander-
son & Tracy. This 200 page tutorial and reference
manual is included with MasterFORTH.

• Floating Point & HIRES options available.

• MasterFORTH - S100.00 APPLE & CP/M; S125.00
Macintosh & IBM PC. Floating Point & HIRES -
$40.00 each.

• Publications

• MASTERING FORTH - S18.00

• 83 International Standard - $1 5.00

• FORTH-83 Source Listing 6502,Z-80,8086 - S20.ea.

Contact:

MicroMotion

12077 Wilshire Blvd.. Ste. 506

Los Angeles, CA 90025

(213) 821-4340

CIRCLE 73 ON READER SERVICE CARD

can be modified or optimized as desired.
No optimizer is included with the
package.

Data types handled by LC include char,
inl, unsigned im, and pointers to any of
these. Arrays are permitted but are lim-
ited to only one dimension. Pointer arith-
metic is in exact accord with K&R as are
most of the library functions. The lack of
floating point capability is addressed by a
special library that provides interfaces to
the floating point code in the TRS-80's
ROM. This makes number crunching pos-

sible, but such programs are not portable.

Since LC is a subset, only the Sieve and
integer-sort benchmarks are applicable to
it. Table 5 (page 102) presents the results
(measurements were made on an early-
production model 4 operating in model 3
mode at a clock speed of 2.2 MHz).

The absence of structs looks to us to be
only a minor inconvenience (the other
omissions have posed no problems in my
applications). Programs created with it
include a virtual-memory full-screen edi-
tor, a communications program including
multiple file-transfer protocols, and many

Dr. Dobbs says,

WE HAVE
GOOD NEWS

• ••

"WINDOWS FOR C can offer you a convenient, reliable means
of implementing windows in your text handling programs

•'..video output is fast and clean...

"..documented in well-written and readable English...

.has obviously been well thought out and very cleanly executed.

"WINDOWS FOR C is a professionally-oriented set of
programming tools. ..that we can recommend."

— Ian Ashdown (Dr. Dobb's Journal, November 1984)

MORE THAN A WINDOW DISPLAY SYSTEM, WINDOWS FOR C is
a video toolkit for all screen management tasks: pop-up menus and
help files, status line, keyboard interpreter, word wrap, auto scroll,
highlighting, color control, overlay and restore, plus a library of over
50 building block subroutines.

EASY TO LEARN. EASY TO USE. Once you've tried WINDOWS
FOR C, you'll wonder how you ever managed without it.

Full support for IBM PC/XT/AT and compatibles, plus interfaces for non-IBM
computers; Lattice C, CI-C86, Mark Wm. C, Aztec C, Microsoft C, DeSmet C
(PC/MSDOS).

The Good News .

- ■--!

WINDOWS FOR G

ADVANCED SCREEN MANAGEMENT MADE EASY

A Professional Software Tool From

CREATIVE SOLUTIONS • 802 . 848-7738

2 1 Elm Ave., Box L2 • Richford , VT 05476

WINDOWS FOR C (specify compiler and version) $ ! 95
Demo disk and manual (applies toward purchase) 130

MasterCard & Visa Accepted

Shipping $250

VT residents add 4"u tax

small utilities. One of the best features of
LC is the ease with which its libraries may
be modified; possibly the poorest parts of
it are the lack of optimization, the failure
to perform constant expression evaluation
at compile time, and the absence of
parameterized tidefine commands.

Whitesmiths footnote

Unfortunately, we were not able to obtain
for review a current version of the White-
smiths C compiler. By the time you read
this review. Whitesmiths should be ship-
ping version 3.0. which claims to improve
some of the problems wc encountered
when we analyzed an older version (ver-
sion 2.2) of the compiler.

If you're considering the puchase of
this older version of the Whitesmith 's
compiler, you should be aware of the fol-
lowing concerns: installation is poten-
tially difficult because files are auto-
matically erased during a relatively long
link session. You'll need a hard disk to
run the supplied DOS 2.0 batch files.
Also, the two manuals supplied do not
cover certain MS-DOS implementation
issues very well (e.g., no installation pro-
cedure for floppy-based systems running
DOS 2.0 or 3.0), and the benchmarks
used in this review had to be modified for
them to compile correctly, due to the lack
of automatic initializers. However, if you
do cross development work with mini-
computers and MS-DOS equipment, this
version of Whitesmiths may be the right
choice.

That's it!

A package is available for just about any
requirement. Most of the packages are
good, solid products. We have tried to
point out the few that seemed less than
adequate. Since everyone has different
needs, it is difficult to make a hard
recommendation.

The C programming language con-
tinues to evolve. Originating as a mini-
computer systems programming lan-
guage, it is now beginning to appear as a
serious contender in the microcomputer
arena. Is C a reality on microcomputers?
The answer to that question is
undoubtedly "Yes!"

References

1 . Clark. David D. "Two More Versions of C
forCP/M."flrT£(May 1984): 246,

2. Pournelle, Jerry. "A BASIC and Pascal
Benchmark, Elegance, Apologies and
FORTH." BYTE (Oct. 1982): 254.

Authors Steve Leibson, Fred Pfahter, and
Jim Reed can be reached ax CadnetLx
Corp. , 5757 Central Ave. , Boulder, Colo.
80301.

CIRCLE 28 ON READER SERVICE CARD

96 COMPUTER LANGUAGE! FEBRUARY 1985

ADVANTAGE # 1

At Programmer's Connection, we deliver the latest
versions of products and keep you informed about
upcoming "new releases" so that you can make
the best purchasing decision. We care about the
value you receive for your dollar, and that
means getting the newest version of a
product with the latest enhancements.
When you are expecting a new release,
Programmer's Connection delivers.

Discover the advantages of buying
from Programmer's Connection:

1. We offer the latest version of a product.

2. Most popular products are in stock ready to be shipped.

3. Receive same manufacturer's support as if buying direct.

4. Experienced professional programmers are on stall.

5. Select from over 200 of the best software products available.

6. Knowledgeable and courteous sales staff.

7. Significant discounts off of retail prices

8. No extra charge on prepaid orders, including major credit cards

9. Reasonable charges for shipping and handling
10. Toll free service from Canada and the Continental U.S.

Programmer's Development Tools:

Computer Innovations C-86 395 309

DeSmet C Compiler with Debugger 159 145

Lattice C Compiler 500 295

Mark Williams C Compiler w/Source

Debugger 500 449

Wizard C 450 409

Xenix Development System by SCO 1350 1099

C Interpreter ■ RUN/C

t way to learn the C language. In addit
jse this system to develop and debug
compilation.

C UTILITIES:

Phact by Phact Associates 395 359

Windows For C by Creative Solutions 195 159

OTHER UTILITIES:

OTHER LANGUAGES:

8088 Assembler w/Z-80 Translator 2500 AD . 100 89

BetterBASIC by Summit Software Sale! 200 149

Janus/ADA + Tools by R&R 700 499

Modula-2/86 by Logitech 495 439

Professional BASIC by Morgan Computing . 95 89

STSC APL*Plus/PC

New version 4.0 of this powerful APL development
system is now available. Call us for product details.

Sale price $469, List price $595.

C UTILITIES:

Btrieve by SoftCraft 250 194

Communications Library by Greenleaf 160 124

C Power Paks from Software Horizons CALL CALL

C-Tree by Faircom 395 359

C To dBase by Computer Innovations 150 139

C Utility Library by Essential Software 149 119

ESP for C by Bellesoft 349 319

Graphic by Scientific Endeavors 195 169

Greenleaf C Functions Library 175 129

Halo Color Graphics 200 125

CIRCLE 84 ON

Dr. Halo by Lifeboat 95 79

Periscope debugger by Data Base Decisions 295 269

Pfix-86 Plus by Phoenix Software 395 299

Plink-86 Overlay Linkage Editor 395 299

Pmate text editor by Phoenix Software 225 159

Polytron Products CALL CALL

Profiler by DWB & Associates New Low Price! 125 99

Screen Sculptor by Software Bottling 125 109

CodeSmith-86 Symbolic Debugger

by Visual Age

The New Version 1.9 now supports In-Line Code Assembly.
Save S26 off Manufacturers List Price of $145.

Special Price $119

Prices are subject to change without notice.

Call for our Catalog consisting of 200 +
Programmer's Development Tools Exclusively
^^ , for IBM PC's and Compatibles, jmm>.

-.~ij.'-~. . Account is charged when order is shipped. ^9

In Canada:

1-800-336-1166 1-800-225-1166

Programmer's Connection

136 Sunnyside Street
Hartville, Ohio 44632
(216)877-3781 (In Ohio)
"Programmers Serving Programmers"

READER SERVICE CARD

Announcing a

TOTAL PARSER GENERATOR

SLICE YOUR COMPILER
DEVELOPMENT TIME

An LR{1) parser generator and several sample compilers,
all in Pascal for your microcomputer.

• Generates parser, lexical analyzer and skeleton
semantics

• Universal, error recovery system

• Adaptable to other languages and computers

• Interactive debugging support

• Thorough documentation

• TURBO PASCAL'" INCLUDED FREE OF CHARGE

• Includes mini-Pascal compiler, assembler, simulator
in SOURCE

SPECIAL INTRODUCTORY OFFER S 995

QPAR5ER'" rjns on IBM. PG'DOS ,n Turoo Pj«ai P*r*j?i -jrnroi&r n oCme, Form, a n «| S e m lource
OPAFISEFItAki'saijia.Tiii^'andm-ngiaipja tvo'iiti como-wc fig-h-DeftamAnesicwTipim wtifi»kp!a.
ion iem»nrics m P»cU soutcs Easy rn add 1 nil semanifcs r« YOUR afjrr.K.-tr.an Excellent !r_< indutlTrial
»nd academic use An accompdnyinc Kntuoo* [SR6 f^isii"^; avfltlfHe in 1905 tramfig can M
nrrangad Dernoniih. avotkaolo jfir S&0

EducaliDnal arid quantity flisccunts avaiiarjfe Cr*c*. mcmey qrrjer. Uaslccaid. Visa CaliWn.a fes:-

WRITE OR CALL FOP FREE BROCHURE
Technical details: call J08/255-5574. immediate delivery Call TODAY'

SYSTEMS, INC.

H6J Hyde Ave. San Jose CA 95129

TOLL FREE: 800-538-9787

iCaJ.forrva i &sidenis call JOB/?55-5S7d}
Turbo Pascal is a registered trademark dI Borland international

CIRCLE 23 ON READER SERVICE CARD

C Source Code

RED

Full Screen Text Editor

IBM PC, Kaypro, CP/M 80 and CP/M 68K systems.

■ RED is fas:! RED uses all of
your terminal's special func-
tions for best screen response,
RLD handles files a; large as
your disk automatically and
quickly.

• RED is easy to use for writers
or programmers. RED's com-
mands are in plain English.

• RED comes with complete
source code in standard C.
RED has been ported to main-
frames, minis and micros.

• RED comes with a Reference
Card and a Reference Manual
thai provides everything you
need to use RED immediately.

• RED is unconditionally
guaranteed. If Tor any reason
you are not satisfied with RED
your money will be refunded
promptly.

RED: S95
Manual: $10

edward k ream

Call or write today for

for more information:

Edward K. Ream

1850 Summit Avenue

Madison, \VI 53705

(«B) 231-2952
To order:

Hither tile UUS C compiler or the Aztec Cll compiler is required Tor OVM80
systems. Dieilai Research C compiler s 1.1 is required for CP/M 6HK svstents. No
compiler is required lor IBM or Kaypro systems.

Specify both the machine desired (IBM. Kaypra or CP/M) and the disk format
described 1 8 inch CP/M single density or exact type of 5'/j inch disk I.
Send a check or money order lor S95 ($105 U.S. lor foreign orders). Sorrv. I do
NOT accept phone, credit card, or COD orders. Please do nol send purchase orders
unless a check is included. Votir order will be mailed lo you within one neck.

Dealer inquiries invited.

Ulhat bJsb cnuld uou do P If you had s
3-D GRAPHICS LFilUGUnEE to harness the
camputat isna I pourer af idour cnmputEr?

E.G. is b new programming language that
provides mast or the computing pntuer of
a dedicated CFiD 5Y5TEm at en incredible
pricei less than one cent an the dollar

EdB.is a complete graphics language:
J automate the generation of drawings
j design math intansivE products

* unurap S. template curved surfaces

* output ascti files to drive n.c.tnals

* accuracy exceeds 1 part in 35, BBB

43 cDmronds control leaping, branching
disc access, internal data structures,
suttrcjut ines and output devices.

E.G. is also a poujerFul past-processor:

-r- plat your numerical data in 3-D

J virtually unlimited macro capability

* T J e7J,l E J "turn-key" applications

* 10BB data paints, 1BB0 lines per file

* optional "best fit" plot algorithm

Easy to use, rather than easy to learn

E.E.providEs the programmer ujith every
niathemat lea I aperat Ian £. Funct inn Found
n emii i C + pmuerFul, proprietary shape
generating and manipulating commands .

_,,_„ EUCLID GRAPHICS

BVSIEms developed at mnmuFiL $ E5

E4K APPLE SE^EN 5EFI5 SDFTlunRE CDuTpCete SEBC

D.FJ.5 3.3 35 Cape George Wye
Part Taiuensend. Da,
98363
<£06> 385-3771

Free updates
For one year

signed nondisclosure agreement required

CIRCLE 4 ON READER SERVICE CARD

BUFFEHED i/O BOARD Introductory Price 'S59.95

Withdespool [unctions, protocols supported: XONIXOFF, ETX: ETBIACK

£>

Ml.

tt t IM

Serial 110
110.19200 Baud

Serial I tO
110-19200 Baud

I < B r>

Parallel
In

<£>j

Parallel
Out

Programmable .
Timer

ZB0 Conlrql
Processor

B4K

Bultei RAM

Vectored
Interrupts

, S-100 Interface
16 Bit I/O Addressing

2K Program EPROM

Programmable Functions;
Dynamic Buttering

BO CHARACTER VIDEO BOARD

25 Lines with slatus, compatible with Wordstar & dBase

* $49.50

+ 44

1 * T <">

Sep,
Sync.

Composite
Video

2716 Char.
Gflneratot
EPROM

2716 Allemale
Character
Gen. EPROM

Bell

Out

Keyooarc

Parallel

Port

1 1—

- 1 ' r

B275 CRTC
Reverse Video
Highlight, Blink

ZS0 Control
Processor

Type Ahead
Bulfer

t L

, i

Screen
RAM

2K Program
EPROM

S-1G0 Interlace
fi Bit I/O
Addressing

Memory
Mapped Iro

Includes Bareboard, Healsink & Documentation Call or write for more intormation,

P.O. S*x60t KIP

!♦#«&*. U. 60195 aMi
(3iZ) 359-7337

OEM dealer pricing available. $3.00 S/H. it. Res. add 7% tax l^ff ^~^

davta" - al Aihtrsn - r.l* Curp. — Wordstar" - al Miciapro Im'i. Ctiwp ^^^p

CIRCLE 70 ON READER SERVICE CARD

COMPUTER LANGUAGE ■ FEBRUARY 19B5

Basic data types

Lattice

yes no

yes yes

yes no

yes yes yes

no yes yes yes yes yes

Microsoft yes no
(2.03)

yes yes

yes no

Microsoft yes yes yes yes

(3.0)

Mark yes yes

Williams

yes yes

= Systems

? Supersofl yes yes yes yes yes no no no no no yes

C Ware yes no yes yes no no

yes no

yes yes yes

Datalight

yes

yes no

Ecosoft

yes

yes yes

yes

no no

yes no no no

Ale

yes yes

BDS

yes yes

Code yes yes

Works

yes yes

no no

Software yes yes yes yes

Toolworks

yes

yes yes

Hendrix yes no

yes no

no no

no no no no

Micro ware yes yes yes yes yes yes yes yes yes yes yes

Introl yes (default) yes yes yes yes yes yes yes no yes no

(unsigned)

yes yes yes no no yes yes yes yes yes

yes yes yes yes yes yes yes yes yes yes yes yes yes yes

5. Rational yes yes yes yes yes yes yes yes yes yes yes no yes yes yes yes no yes

■5 Wizard yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes

^ Systems

{ Computer yes yes yes yes yes yes yes yes yes yes yes no no yes yes yes no yes

t Innovations

yes yes no yes no yes

£ Manx yes yes yes no yes no yes no yes yes yes no no yes yes yes no yes

Software
Systems

C-Systems yes no yes yes yes no yes no yes yes yes no no yes yes yes yes yes

yes yes yes yes no yes

g Digital yes yes

S Research

yes yes

yes yes yes yes

no yes

yes yes yes yes yes yes yes yes yes yes
yes yes yes yes yes yes yes yes yes yes yes yes yes yes

yes yes yes yes yes yes yes yes

no no yes yes yes yes yes yes no yes

yes no yes yes yes yes yes yes yes yes

no no

yes yes yes no yes

yes

"One dimensional. ** Is a library function.

Table 3.

Library functions

Lattice

yes

Microsoft 2.0

yes

Microsoft 3.0

yes

Mark Williams

yes

"5.

Rational

yes

E
°

Systems

</l

Wizard

yes

Systems

Computer

yes

Innovations

Supersoft

yes

Manx Software

yes

Systems

C-Syslems

yes

CWare

yes

Dalolight

yes

Ecosoft

yes

'5.

Aleor

yes

BDS

yes

Code Works

yes

Software

yes

Toolworks

ftertdrix

yes

Microware

yes

Introl

yes

-o

Digital

yes

Research

yes
yes
yes
yes
yes

yes

yes
yes

yes
yes
yes
yes
yes
yes
yes
yes

yes
yes
yes
yes

yes
yes
yes
yes
yes

yes

yes
yes

yes
yes
yes
yes
yes
yes
yes
yes

yes
yes
yes

yes

yes
yes
yes
yes
yes

yes

yes
yes
yes
yes
yes
yes
yes
yes

yes
yes
yes
yes

yes

no
yes

yes
yes

yes

no
yes
no
yes

yes

S10

yes

external

yes

Table 4.

100

COMPUTER LANGUAGE ■ f EBRUARY 1985

C ENGLISH.

The C Generation Language.

WhatiscENGUSH?cENGLISH is a comprehensive fourth generation
procedural language based on dBASE IP syntax, it is portable to a
wide range of micros and minis. The language features user-trans-
parent interfaces to a wide range of popular C compilers, operating
systems, and data base managers.

How is portability achieved? cENGLISH through its compiler inter-
face translates cENGLISH into documented C source and uses a host
C compiler to produce native machine code.

CfNGUSH
SOURCE

EXECUTABLE

NATM

MACHINE COOE

•

SOURCf

Csourca can bo 9mtMdd4ti IncEHGL&H sourcs.

Differences in the operating system and data base manager are
handled by the runtime libraries.

The result is that cENGLISH source can be compiled without modi-
fication on any micro or mini configuration supporting cENGLISH.
What about performance? cENGLISH executes FAST, just like any
compiled C program.

How easy is cENGLISH to use? While cENGLISH is a powerful high
level language that can accommodate complex software develop-
ment, it remains simple and straightforward to use.

Call or write for availability of c ENGLISH for the following configu-
rations-
Compilers:

Standard O/S compilers: Lattice C" for MS/DOS'"
Operating Systems:

UNIxf UNIX-like, MS/DOS: Coherent: VMS"
Data Base Managers:

C-ISAM" and INFORMIX: UNIFY," ORACLE: PHACT Logix"
Foreign Language Versions:

German, French, Spanish

Atterrfion MS/DOS users. Demo version and special introductory offer
available for IBM PC: XT,' AT," and other MS/DOS systems.
Requirements: 256K. hard disk or two floppy disk drives, and
MS/DOS 2.1 or higher.

Attention dBASE II and dBASE III users. dBASE II to cENGLISH
Converter now available; dBASE 111 Converter available later this
quarter. Converted code is portable to micros or minis and executes
as fast as original cENGLISH source.

l&DOStiaTOWT«*o»Mcf3Krt IncCcrtfirPnotTaawrortrfMcrtWllnrrearrparT, ^VSr*o?TjdencnofDs|t3l£juiprwn!
CapMwn C^SAMaroWWWUMfiwimorWtf 6#cwa0awtiiB»$ni«Ta JK OocrtsonacwartaOrKMlnc FMAC1

q S tTOC*TO* 0< RXJCt AMOOOHS I 0C« * P»OC*KJft <* LO0«l SO«W» »X *W PC XT d*3 H m tTM«mc«1UC«lnWnono<K*
BknKMUMQcrrosCwpoarwn IMPf n a HdHnodl oUMV tap

SAMPLE cENGLISH PROGRAM

IDENTIFICATIONS

MODULE Mmirro™

AUTHOR DCs

DATE 8/29/64

REMARKS SamptecENGLISH program thai odds lirst
names 'a o f ile
END IDENTIFICATIONS

GLOBALS

FIXED LENGTH Ions

FIXED LENGTH 15 fnome
END GLOBALS

MAIN PROGRAM

BEGIN
CLEAR SCREEN
SET ECHO OFF

USE ■NAMES"

VIEW BY TD_FNAME " ASCENDING

AT 23,1 SAY' Add record? Yor N"
AT 23.25 ENTER ons USING "' '

WHILE onsEO'V
CLEAR GETS

AT 6. 1 SAY "Enter first name"
AT 6.20 GET Fnome
READ SCREEN

INSERT

Frame = Frame
END INSERT

AT 12, 1 SAY Welcome to cENGLISH; & Fnome

WAIT

AT 14, 10 SAY "HIT ANY KEYTO CONTINUE"

STORE" "TO Fnome

STORE" "TOons

AT 23. 1 SAY "Add onolhet record'' ¥ or N"

AT 23.30 ENTER ons USING" 1 '

CLEAR ROW 1THRU 23

END WHILE

AT 1 2, 10 SAY "That's oil lor raw'"
UNUSE "NAMES "
SET ECHO ON

END PROGRAM

** M Hklt

I'd like to know more about cENGLISH.
Please send further information.

ftwTNiime

Til If

Company

"■ ■ r " ■

Addrew

Ciry
Check one:

Suit
1 Dealer

Syiiem House
. .TCit Ontario, Chicago, IL
Tetex 510315 Phone (3 1 2| 9444MG

] Distributor

]r i in^a cUNFi Canada, Inc. Complete La Laurendenne,
"i5 r _

42 5 St. Amable, Suite 1 65, Quebec, Canada G 1 R5E4
Phone |416| 524 4641

:,(JanadalilK5t4

A4QS4 I

CIRCLE 26 ON READER SERVICE CARD

101

TRS-80 benchmark comparisons

Source code size (bytes)
Assembly code size (bytes)
Executable code size (bytes)

Compile time (min:sec)
Assembly time (min:sec)'

Execution time (min:sec) 2

Sieve

390
2,768
3,129

00:28
02:53

01:52

Mysort

296
2,750

3,139

00:28
02:59

00:32

1 . EDAS is disk bound when compiling LC-generated code; these times reflect perfor-
mance of a two-pass assembler taking all input from floppy disk files.

2. Times include total program load and run time. Actual sorting time on mysart was
about 9 sec; remaining time includes array initialization and results reporting.

TableS.

The C Interpreter:

nstant-C

Programming in C has never been Faster.
Learning C will never be Easier,

Instant-C " is an optimizing interpreter for the C language that can make
programming in C three or more times faster than using old-fashioned
compilers and loaders. The interpreter environment makes C as easy to
use and learn as Basic. Yet lnstant-C m is 20 to 50 times faster than inter-
preted Basic. This new interactive development environment gives you:

Instant Editing. The full-screen editor is built into Instant-C™ for imme-
diate use. You don't wait for a separate editor program to start up.
Instant Error Correction. You can check syntax in the editor Each error
message is displayed on the screen with the cursor set to the trouble
spot, ready for your correction. Errors are reported clearly, by the editor,
and only one at a time.

Instant Execution. Instant-C ''' uses no assembler or loader. You can
execute your program as soon as you finish editing.

Instant Testing. You can immediately execute any C statement or func-
tion, set variables, or evaluate expressions. Your results are displayed
automatically.

Instant Symbolic Debugging. Watch execution by single statement
stepping. Debugging features are built-in; you don't need to recompile or
reload using special options.

Instant Loading, Directly generates .EXE or .CMD files at your request

to create stand-alone versions of your programs.

Instant Floating Point. Uses 8087" co-processor if present.

Instant Compatibility. Follows K & R standards. Comprehensive stand-
ard library provided, with source code.

Instant Satisfaction. Get more done, faster, with better results.
Instant-C™ is available now, and works under PC-DOS, MS-DOS* and
CP/M-86*.

Find out how Instant-C™ is changing the way that programming is done.
Instant-C™ is $500. Call or write for more information.

Rational (617)6535194

AvauvjAiai pQ Box 480

Systems, Inc. Natick, Mass. 01760

Trademarks: MS-DOS (Mcrosott Corp.), B087 (Inteil Corp ), CP/M-86 (Digiial Research. Inc ), Insant-C (Rational Systems, Inq )

Manufacturer Product

Alcor AkorC

1 1 32 Commerce
Richardson, Texas 7508 1
(214)238-8554

BD Software BDSC

P.O. Box 2368
Cambridge, Mass. 02238
(617)576-3828

C-Systems C

P.O. Box 3253
Fullerton, Calif. 92634
(714) 637-5362

CWare DeSmelC

P.O. Box C

Sunnyvale, Calif. 94087

(408) 720-9696

The Code Works Q/C

Box 6905

Santa Barbara, Calif. 93 1 60

(805)683-1585

Computer Innovations Optimizing C86

980 Shrewsbury Ave.
Tinton Falls, NJ. 07724
(201)542-5920

Datalight Small-C

1 1557 8th Ave. N.E.
Seattle, Wash. 98125
(206)367.1803

Digital Research C

60 Garden Cl. Rd.
P.O. BoxDRI
Monterey, Calif. 93942
(408) 646-6464

Ecosoft Eco-C

6413 N. College Ave.
Indianaplolis, Ind. 46220
(317)255-6476

Jim E. Hendrix Small-C

P.O. Box 8378
University, Miss. 38677
(601) 234-7508 eves.

Introl C

647 W. Virginia St.
Milwaukee, Wis. 53204
(414)276-2937

Lattice C

P.O. Box 3072

Glen Ellyn, III. 60138
(312)858-7950

Manx Software Systems Aztec C

P.O. Box 51

Shrewsbury, N.J. 07701
(800)221-0440

Mark Williams Co. MWC86

1 430 W. Wrightwood Ave.
Chicago, III. 60614
(800)692-1700

Microsoft C

P.O. Box 97200
Bellvue, Wash. 98009
(206) 828-8039

MiSosys Inc.

P.O. Box 239

Sterling, Va. 22170-0239

(703)450-4181

Rational Systems Instant C

P.O. Box 480
Natick, Mass. 01 760
(617|653-6194

Software Toolworks C/80

1 5233 Ventura Blvd. Ste. 1118
Sherman Oaks, Calif. 91 403
1818)986-4885

Supersoft C

P.O. Box 1628
Champaign, 111.61820
(217)359-2112

CIRCLE 56 ON READER SERVICE CARD

102 COMPUTER LANGUAGE ■ FEBRUARY 1985

Wizard Systems
11 Willow Ct.
Arlington, Mass. 02174
(617)641-2379

Wizard C

STRUCTURE FOR BASIC

u . . . PROFESS! ONAL PROGRAMMI NO
ENVIRONMENT

ELIMINATES

LINE

NUMBERS

ALLOWS
MULT! -LINE
CONDITIONALS

PCDOS/MSCOS

WORKS WITH

BASICS

1 NTER FRETEfi

COM PI LXR

FULL ERROR
LOGGING

PROGRAM
U STER

1 BENDORF

i206 S. MAIN

p.o. box 5sia

ROSWELL, NM

S8201
505 347-5701

UISA/MASTERCARD

LABELED

PROCEDURES

MACROS

SUB-ROUTINES

LIBRARIES

Users'
Group

Over 40 volumes of public
domain software including:

• compilers

• editors

• text formatters

• communications

packages

• many UNIX-like tools

Write or call for more details
The C Users' Group

415 E. Euclid* Box 97

McPherson, KS 67460

1316)241-1065

A general purpose programming
language for string and list
processing and all forms of

non-numerical compulation.

SNOBOL4+ -The entire
SNOBOL4 language with its superb pattern-matching
facil ities • Strings over 32.000 bytes in length • Integer
and floating point using 8087 of supplied emulator *rA
■ ASCII, binary, sequential and rondom-
access I 'O • Assembly Language inter- j(r

to 300K bytes js^* 4 J

bam «*r\ t\A VJ

grams and functions

.g. T _

Forall8086 66 PC MS-DOS or

CP M-36 systems. 12BK minimum

D5DD specify DOS CFM formal

Send check VISA. M'Cto:

'°Catspaw, Inc.

P.O.Box 1123-Salido. CO81301

$95

plus'3s/h
'303 53 1 * 3BBJ

CIRCLE 10 ON READER SERVICE CARD

CIRCLE 43 ON READER SERVICE CARD

CIRCLE 9 ON READER SERVICE CARD

* * * EASY TO USE # * *
Macro Programs for

■

H^^^fflf I

1 1'

NOW tor PC-DOS

::•

i i

# i ■*
i *

Customized Tiblv* -RAMd Ilk with .BAT II la.

Toggle extended character iet and prinipr
control.

i *■

# t
i h

Our old standby, DearJohn, mailing Hal
management v»tth u us lilts. Sill! $$7,50 CP/ M or
PC-DOS

" 4

i •*

• :»

: i

"NEW* All You Wanted lo know abour
Spellbinder but Couldn't Find Out". A
programmer 1 ! notebook of major command*
and how to use. $10.00 pef copy

;;•

We have a Iqng Hit qI CJUflflfll mpcroi lor me In
the office and by the programmer. Send tl .00
for catalog.

;;'•

• »

<«0mPUTEn RESOURCES _.

" *

lj»

P.O. Box 1569 Kamuela, Hawaii 96743
[808} 885-7905

isj

ADVERTISE

in the April issue of

COMPUTER
LANGUAGE

Special distribution

at the

West Coast

Computer Faire

Reservation deadline:

February 6th
Call (415) 957 9353

BDSCv1.5

For FAST
Development

• Fastest compile to execute
cycle of any CP/M 80®
Compiler.

• Dynamic
debugger,

• 180 page
manual.

TERMS: check, c.o.d., charge card. T

{316)431-0018

KS res. add 4% tax.

$120.00

Plus $2.50

Shipping 6 handling

Dedicated Micro Systems Inc

P.O. Box 481 Chanute, KS 66720 o

CIRCLE 14 ON READER SERVICE CARD

CIRCLE 21 ON READER SERVICE CARD

c Programmers

B-Trees

For

75.

1+2.00 Postage

source code included

The Softfocus B-Trees record index-
ing library will help you develop
sophisticated application programs.
With Softfocus B-Trees you get:

• high speed file handling for up
to 16.7 million records per file

• customizable BDS or K & R
standard C source code

• no royalties on applications
programs

• support random and sequential
file access

• includes example programs

« Softfocus
1277 Pallatine Dr.. Oakville. ont.
Canada L6H 1Z1 (4161 844-2610 i

CIRCLE 79 ON READER SERVICE CARD

Com pu Pro
Users Group

Over 700 Members
and Growing!

Join now to receive:

- Technical Newsletter

- Bulletin Board & RCPM

- Software Reviews

- New Product Announcements

- Hardware Help

- And More!

All for just $24/year!

Send check to:

C-PRO Users Group

P.O. Box 2146
Woodbridge, VA 22193

CIRCLE 61 ON READER SERVICE CARD

-+ All l&JbS. DftlA PHODOCTS ■ *>
IS-Blt. s-ioo But: Systomu & Boards

• * « FIREBALL DISCOUffTS - ' -

25% off List Price (UPS extra) tor

Cashlefs Check or Money Order

+ + +

You save MONEY = We save TIME

♦ + +

See -+ LOMAS 4- ad In bvte

or request complete specifications

\ UBHTNING SPEED /
8028B-CPU a B02B7-NOP: <tr<68. 75
; I THUNDERING PERFORMANCE I I

aoiee & i/o a 256K-dram * 5.25-/8-
disk controller 4 CCP/M-86: $1I»6.2B

Configured for SmnTuc-STOO printer:

T»H>Vid9O-GA970C VOT: $1121.75

Spellbind f -5. 30 W/P: $371.26

HIERATIC COMPUTE* SYSTEMS

■OX 133l MEDFOKD. I-IA 03155

[6171 683-6540

CIRCLE 45 ON READER SERVICE CARD

103

ADVERTISER INDEX

Alcor Systems. .
Amber Systems.
BD Software .. .

PAGE

NO.

4
42

CIRCLE
NO.

... 1
... 2
... 3

Bendorf & Associates 1 03 10

Blaise Computing Inc 83 11

Borland International 6,7 6

Business Utility Software 86 17

C Journal 86 5

C Systems 59 16

C Users Group 1 03 43

CWare 38 18

C-Linelnc 101 26

C Pro Newsletter 103 61

Catspaw Inc 103 9

Code Works (The) 24 42

CompuPro COVER IV 12

CompuServe 2 7

Computer Innovations 41 8

Computer Resources of Waimea 1 03 14

Computingl 74 37

Compuview Products 81 40

Creative Solutions 96 28

DWB Associates 23 20

Data Base Decisions 47 30

Datalight 24 19

David Smith Software Co 90 36

Dedicated Micro Systems Inc 103 21

Ecosoft 90 22

Essential Software Inc 32 27

(LISP) FOR A.I.

UO-LISP Programming Environment

The Powerful Implementation of LISP

for MICRO COMPUTERS

LEARN LISP System (LLS.l)
(see description below €39 95

UO-LISP Programming Environment
Base Line System (BLS.l) $49.95

Includes: Interpreter, Compiler,

Structure Editor, Extended Numbers,

Trace, Pretty Print, various Utilities,

and Manual with Usage Examples.

(BLS.l) expands to support full system

and products described below.

UO-LISP Programming Environment: The Usual LISP Interpreter Functions,
Data Types and Extensions. Structure & Screen Editors. Compiler, Optimizer, L1SP&
Assembly Code Intermixing, Compiled Code Library Loader, I/O Support. Macros,
Debug Tools, Sort & Merge, On-Une Help. Other Utility Packages. Hardware and
Operating System Access. Session Freeze and Restart. Manual with Examptesexpands to
over 350 pages. Other UG LISP products include: LISPTEX text formatter, LITTLE
META translator writing system, RL1SP high level language, NLARGE algebra system.
Prices vary with configurations beyond (BLS.II please send /or FREE catalog.
LEARN LISP System (LLS.l): Complete with LISP Tutorial Guide, Editor Tutorial
Guide, System Manual with Examples, Full LISP Interpreter, On-Line Help and other
Utilities. LEARN LISP fundamentals and programming tec hniques rapidly and effectively.
This system does riot permit expansion to include the compiler and other products listed
above,

LISP Tutorial Support (LTS.l): Includes LISP and Structure Editor Tutonal
Guides. On line Help, and History Loop. This option adds a valuable learning too! to the
UO-USP Programming Environment (BLS.]| Order (LTS.l) for $19.95.
REQUIRES: UO LISP Products run on most Z80 computers with CR'M, TRSDOS or
TRSDOS compatible operating systems. The 8086 version available soon.
TO ORDER: Send Name. Address. Phone No , Computer Type. Disk Format Type. Package
Pnce, 6.5% Tax (CA residents only). Ship & Handle fee of 53.00 inside US. & CN, S10 outside
U.S., Check, Money Order, VISA and MasterCard accepted. With Credit Card include exp. date.
Other configurations and products are ordered thru our FREE catalog.

Northwest Computer Algorithms

P.O. Box 90995, Long Beach, CA 90809 (213) 4261893

FairCom 91

Greenleaf Software 88

HSC, Inc 92

Harvard Softworks 85

Haventree Software 12

Hieratic Computer Systems 1 03

Intro! Corp 62

Korsmeyer Electronics Design 78 34

.29

.44
31
47

.41
.45
.32

Laboratory Microsystems Inc 91

Lahey Computer Systems Inc 94

Lattice Inc 64

Lifeboat Associates 52,53

Lomas Data Products 25

mbp Software & Systems Technology .... 10

MEF Environmental 89

MEF Environmental 86

Manx Software Systems 8

Megamax, Inc 70

Mendocino Software Co. Inc 70

Micro Methods 70

Micromotion 95

MicroTec Research Inc 76

MicroWay 72

Mindbank Inc 22

Next Generation Systems 66

Northwest Computer Algorythms 104

Oasys 63

Oasys 65

Opt-Tech Data Processing 86

Opt-Tech Data Processing 95

Parsec Research 95

Phoenix Computer Products Corp 48

Plum Hall 67

Poor Person Software 58

Prentice Hall 16

ProCode 85

Programmer's Connection 97

Programmer's Shop (The) 60

QCAD 98

RR Software COVER II .

Rational Systems, Inc 1 02

Realia, Inc 93

Reom, Edward 98

SLR Systems 72

Seven Seas 98

Shaw American Technologies 58

Simpliway Products Co.

Soft Advances

Soft Craft Inc

Softfocus

Software Horizons

Solution Systems

1 03

Spruce Technology Corp 87

.35

.62

.53

.64

.54

.39

.15

.55

.69

.13

.48

.52

.73

,57

.63

.74

.46

.75

.67

.49

.50

.83

.60

.84

.98

.23

.58

.56

.59
. 4
.92
.70
.77
.78

79
.25
.93

99
.97
.33
.87

COVER III 72

CIRCLE 46 ON READER SERVICE CARD

104 COMPUTER LANGUAGE ■ FEBRUARY 1985

Steve Rank Inc

Stride Micro

Summit Software

Systems Guild

Thunder Software

UniPress

Western Wares

Whitesmiths Ltd

Wizard Systems 70

Workman & Associates 78

Zeducomp 80

The index on this page is provided as a service to our readers. The
publisher does not assume any liability for errors or omissions.

68
18
88
67
68
1

.88
,89
.65

38
.90
.85

86
.68
.91

NO POSTAGE

NECESSARY

IF MAILED

IN THE

UNITED STATES

BUSINESS REPLY CARD

FIRST CLASS PERMIT NO 22*81 SAN FRANCISCO, CA USA

POSTAGE WILL BE PAID BY

COMPUTER

LANGUAGE

PO. BOX 11747
PHILADELPHIA, PA 19101

NO POSTAGE

NECESSARY

IF MA! LEO

IN THE

UNITED STATES

BUSINESS REPLY CARD

FIP.ST CLASS PERMIT NO. 22*81 SAN FRANCISCO. CA USA

POSTAGE WILL BE PAID BY

COMPUTER

LANGUAGE

PO. BOX 11747
PHILADELPHIA, PA 19101

COMPUTER

LANGUAGE

-over

Subscribe to COMPUTER LANGUAGE today for only $24.95-

30% savings off the single copy price.

D Yes, start my Subscription to COMPUTER LANGUAGE today. The cost

is only $24.95 for 1 year (12 issues).
D I want to increase my savings even more — send me 2 years (24

issues) of COMPUTER LANGUAGE for only $39.95.
D Payment enclosed □ Bill me

Name

Company

Address

City, State, Zip .

Please allow 6-8 weeks for delivery of first issue. Foreign orders must be prepaid in U.S.
funds. Canada orders $30.95 per year. Outside the U.S., $36.95/year for surface mail or
$54.95/year for airmail.

LANGUAGE

Subscribe to COMPUTER LANGUAGE today for only $24.95— over

30% savings off the single copy price.

G Yes, start my Subscription to COMPUTER LANGUAGE today. The cost

is only $24.95 for 1 year (1 2 issues).
□ I want to increase my savings even more — send me 2 years (24

issues) of COMPUTER LANGUAGE for only $39.95.
D Payment enclosed D Bill me

Name

Company

Address

City, State, Zip

FREE INFORMATION

NO POSTAGE

NECESSARY

IF MAILED

IN THE

UNITED STATES

BUSINESS REPLY CARD

FIRST CLASS PERMIT NO. 27346 SAN FRANCISCO, CA USA

POSTAGE WILL BE PAID BY

COMPUTER

LANGUAGE

2443 FILLMORE STREET • SUITE 346
SAN FRANCISCO, CA 941 15

NO POSTAGE

NECESSARY

IF MAILED

IN THE

UNITED STATES

Name .

BUSINESS REPLY CARD

FIRST CLASS PERMIT NO. 27346 SAN FRANCISCO, CA USA

POSTAGE WILL BE PAID BY

COMPUTER

LANGUAGE

2443 FILLMORE STREET ■ SUITE 346
SAN FRANCISCO. CA 941 15

Company .
Address —

City, State, Zip

Country Telephone number.

February issue. Not good if moiled after June 30. 1985.
Circle numbera tor which you desire Information.

I obtained (his issue through:

(J Subscription I 1 Passed on by associale

f] Computer Store n Other

Q Retail outlet

Comments .

1 tl 21 31 41 SI 61 7) 81 91 101

2 12 22 32 12 52 62 72 82 92 102

3 13 23 33 43 33 43 73 83 93 103

4 14 24 34 44 54 64 74 84 94 104

5 15 25 35 45 55 65 75 85 95 105

6 16 26 36 44 56 66 76 86 96 106

7 17 27 37 47 57 67 77 87 97 107
B 18 28 38 48 58 68 78 88 98 108

9 19 29 39 49 59 69 79 89 99 109

10 70 30 40 50 60 70 SO 90 100 1 10

111 121

112 122

113 123

114 124

115 125

116 126

117 127
US 126

119 129

120 130

131 141

132 142

133 143

134 K4

135 145

136 146

137 147

138 148

139 149
HO 150

Attn: Reader Service Dept.

FREE INFORMATION

Name .

Company .
Address

City, Stale. Zip

Country Telephone number

February <ssue. Nol good il mailed after June 30, 1985.
Circle numbers for which you desire Information,

I obtained Ihis issue through:

□ Subscription L I Passed on by associate
Tl Computer Store □ Other

□ Retail outlet

Comments.

101

111

121

131

141

10!

112

123

132

142

103

113

123

a :■

143

104

114

124

134

144

105

115

125

135

145

••

106

116

13d

136

146

107

117

127

137

147

108

119

133

138

148

109

119

129

139

149

100

110

120

130

140

150

Attn: Reader 5ervice Depl.

Despite the recent press notices,
multiuser microcomputers aren't

anything new!"

This is the first in a series of
discussions with Hod Coleman,
President of Stride Micro
(formerly Sage Computer) on
the 68000 multiuser market
and its current environment.

0: Why do you say thai?
BC: "The technology to build a
high performance multiuser sys-
tem has been around for five
years. And white some of the
leaders in this industry have been
pretending that micro multiuser
didn't exist, we've been shipping
complete systems for nearly three
years. The benefits of multiuser
are undeniable; it is more cost cf-
feetive, and offers greater flexibil-
ity and utility. But until just re-
cently, the marketing pressure to
be compatible instead oi being
better, has blinded the industry."

0; What do you mean'.'
BC," "Well, for example, the
Motorola 68000 processor intro-
duced 16/32-bit technology to the
personal computer world a long
time ago. It was fully capable of

"A surprising feature is
compatibility. Everybody
talks about it, but nobody

does anything about it."

meeting high performance and
multiuser design requirements in
1980. Instead of this trend taking
off, most energy was spent pro-
moting 8088/8086 products that

were clearly inferior from it tech-
nical point of view. This phenom-
enon leads me to believe that they
will soon rewrite the old proverb:
'Build a better mousetrap and the
world will beat a path to your
door," but only if they can find the
way through the marketing fog."
Q: Are things changing now?
BC: "Yes and no. With the busi-
ness world starling to take more
and more interest in microcompu-
ter solutions, the advantages of a
solid multiuser system couldn't be
kept hidden forever, companies
like ours and a few others were
beginning to make a dent. Instead
of taking a fresh approach, some
of the newest multiuser offerings
will probably only give the tech-
nology an undeserved black eye!
Multiuser is far more than the
ability to plug in more terminals.
It involves things like machine
compatibility, last processors,
adequate memory, large storage
capacities, backup features, net-
working, and operating system
flexibility."

(J: Is this what makes the new
Stride 400 Series different?
BC: "Exactly. Thai sounds self-
serving, but it's true. Today a
number of companies are intro-
ducing their first multiuser sys-
tem. We've been building and
shipping multiuser machines for
almost three years. We know the
pitfalls, we've fallen into some of
them. But we have learned from
our mistakes."
0; Give me some examples.
BC; A hard disk is almost manda-
tory for any large multiuser in-
stallation. Yel, backing up a hard
disk can be a nightmare if you
only have floppies to work with.
That's why we've added a tape
backup opiion lo all the larger
Stride 400 Series machines. It's
irresponsible for a manufacturer
to market a multiuser system
without such backup. Another
good lesson was bus design. We
started with one of our own de-
signs, but learned that it's impor-
tant not only to find a bus that is
powerful, but also one that has
good support and a strong future
lo serve tomorrow's needs. We

'The marketing pressure

to be compatible

instead of being better,

has blinded the industry.'

think the VMEbus is the only de-
sign that meets both criteria and
thus have made it a standard fea-
ture of every Stride 400 Series
machine."

0: What are some of the other
unique features of the 400 Series?
BC: "A surprising feature is com-
patibility. Everybody talks about
it, but nobody does anything
about it. Our systems are com-
pletely compatible with each other
from the 420 model starting at '
$2900. through the 440, on to the
powerful 460 which tops out near
S60.000. Each system can talk to
the others via the standard built-in
local area network. Go ahead and
compare this with others in the in-
dustry. You'll find their little ma-
chines don't talk to their big ones,
or that the networking and multi-
user are incompatible, or that they
have different processors or
operating systems, and so on."
0: When you were still known as
Sage Computer, you had a reputa-
tion for performance, is that still
the case with the new Stride 400
Series?

BC; "Certainly, that's our calling
card: 'Performance By Design.'
Our new systems are actually fas-
ter, our standard processor is a 10
MHz 68000 running with no wait

stales. That gives us a 25% in-
crease over the Sage models.
And, we have a 12 MHz pro-
cessor as an option. Let me add
that speed isn't the only way to
judge performance. I think it is
also measured in our flexibility.
We support a dozen different
operating systems, not just one.
And our systems service a wide
variety of applications from the
garage software developer to the
corporate consumer running high
volume business applications."
(J: Isn't thai the same thing all
manufacturers say in their ads?
BC: "Sure it is. But lo use another
over used-term, 'shop around'.
We like to think of our systems as
■full service 68000 supennicro-
eomputers." Take a look at every-
one else's literature and then
compare. When you examine
cost, performance, flexibility, and
utility, we don't think there's any-
one else in the
race. Maybe
that's why we've
shipped and
installed more
multiuser 68000
systems than
anyone else."

FunncrJy Sjigc Computer

For more information on Stride or
the location of the nearest Stride
Dealer call or write us today.
We'll also send you a free copy of
our 32 page product catalog.

Corporate Offices:
4905 Enerey Way
Reno, NV 89502
(702) 322-6868

Regional Offices:
Boston: (617) 229-6868
Dallas: (214) 392-7070

CIRCLE 72 ON READER SERVICE CARD

THINKING COMPUTERS?

Think CompuPro

........ v

CompuPro •\qplus

"... a coherent, relatively easy to
use [multi-user] system with an
impressive amount of power for
the money and good support...
alternative to stand alone— or
even networked — computers, the
CompuPro 10 has a lot going for

it "

Persona! Computing, 12/84

S7995.00 suggested list witft 40 Mb haid disk

CompuPro 286

A faster version of our multi-user
Model C, of which a recent review
states "... hardware is high quality
... large software base... hundreds
of serious business programs...
versatility and adaptability to
future hardware and software
developments are unmatched..."
Business Computer Systems, 9/84

«V«asyn 1985

$9995.00 suggested list with 40Mb hard disk and cartridge tape
CIRCLE 12 ON READER SERVICE CARD

To arrange a demonstration, call 1-800-367-7816 (in California, 1-415-786-0909).
Viasyn Corporation, 3506 Breakwater Court, Hayward, CA 94545

Internet Archive Audio

Featured

Top

Images

Featured

Top

Software

Featured

Top

Texts

Featured

Top

Video

Featured

Top

Mobile Apps

Browser Extensions

Archive-It Subscription

Save Page Now

Full text of "Computer Language - Issue 06 (1985-02)(CL Publications)(US)"

See other formats